You are on page 1of 328




D. W. TAYLOR, E.D., D. Sc., L.L.D.








Entered at Stationers' Hall. London

Stanhope fl>res
BOSTON. U.S. A 3-2*

THE intention of this work is to treat in a consistent and con-

nected manner, for the use of students, the theory of resistance and
propulsion of vessels and to give methods,, rules and formulae which
may be applied in practice by those who have to deal with such
matters. The contents upon model experiments,
are based largely
such as were initiated in England nearly half a century ago by Mr.
William Froude and are now generally recognized as our most effec-
tive means of investigation in the field of resistance and propulsion.
At the same time care has been taken to point out the limitations
of the model experiment method and the regions where it ceases
to be a reliable guide.

During the years that the author has directed the work of the
U. S. Experimental Model Basin many have
results obtained there

been published in the Transactions of the Society of Naval Archi-

tects and Marine Engineers and elsewhere, so, naturally, the
experiments at the U. S. Model Basin have been made large use of
wherever applicable. It will be found, however, that they are in
substantial agreement with published results of the
the many
work of other experimental establishments of this kind.
Although the coefficients and constants for practical application
are mainly derived from the author's experience at the Model
Basin and elsewhere, and are necessarily general in their nature,
endeavor has been made wherever possible to develop formulae
and methods in such a manner that naval architects and engineers
using the book may, if they wish, adopt their own constants
from their special experience.
For instance, by the methods given it will be found possible to
estimate closely the effective horse-power of a vessel having the
form of what I have called the Standard Series, but it will also be
found possible, by the same methods, to determine with fair accu-

racy the variation of resistance with changes of dimensions, etc.,

of vesselsupon almost any lines for which a naval architect may
have reliable data, and which, on account of satisfactory past

results, or for other reasons, he may wish to use.

The science of Naval Architecture is not yet developed to a point
where our knowledge of resistance and propulsion is complete.
While the author naturally hopes that this volume will at least
partially bridge some of the gaps hitherto existing, much work
remains to be done, and in a number of places attention is called
to the need of further investigation of various questions. While
we know something, for instance, in a qualitative way of the effect
of shallow water upon resistance, information which would enable
many problems arising in this connection
us to solve satisfactorily
islacking, and apparently can be obtained only by much experi-
mental investigation. When dealing with questions of wake and
thrust deduction we are not yet upon firm ground, and it is to be

hoped that the excellent work recently done by Luke in this con-
nection will soon be supplemented by even more extensive investi-

WASHINGTON, D.C., July, 1910.


Preliminary and General














Trials and Their Analysis



The Powering of Ships




i. Stream Lines

1.Assumptions Made. The consideration of stream lines or

lines of flow willbe restricted mainly to the case of the motion of
liquid past a solid. It is sufficient for present purposes to define
a liquid as a fluid which is incompressible, or virtually so, such as
The difficulties in the way of adequate mathematical determi-
nation of the motion of liquids past solids such as ships have
hitherto been found insuperable. The mathematics of the motion
of liquids is complicated; even the simple cases which can be dealt
with mathematically require assumptions which are far from actual
conditions in practice. Thus, when considering the motion of
solids through a liquid, or what is the same thing mathematically,
the motion of a liquid past solids, it is assumed that the liquid
" "
is perfect or has no viscosity and that the solid is frictionless,
that is to say, that the liquid can act upon the solid only by pres-
sure which must at each point be normal to the surface. In most
cases that are dealt with mathematically, it is further assumed
that the fluid or liquid extends to an infinite distance from the

Steady Motion Formula.

2. We cannot deal satisfactorily with
problems of resistance by mathematical analysis, but in spite of
the somewhat assumptions involved, the results of mathe-

matical analysis applied to a perfect liquid are of interest and value

as they indicate tendencies and have large qualitative bearing upon
the phenomena of the motion of water past ships.

One mathematical conclusion in this connection is particularly

valuable. It is known as the steady motion formula and is as
W 2g
z = h .

In the above formula, p denotes pressure of the liquid per unit

area, w denotes weight per unit volume, v denotes velocity of
flow in units of length per second, g acceleration due to gravity
in units of length per second, z denotes height above a fixed level
and h is a constant for each stream line, being called the head.
It isusually convenient to express p in pounds per square foot, W
in pounds per cubic foot, v and g in feet per second, z and h in feet.
The above formula applies to the steady motion of an infinite
mass of perfect liquid. For such liquid the value of h is constant
for all particles passing a point fixed in the liquid. These particles
form a continuous line called a stream line, and in steady motion, no
matter how many twists and turns the stream line takes, the above
formula applies to its pressure, velocity and elevation at every point.
It will be observed that contrary to what might at first be thought,
the greater the velocity at a point of the stream line the less the

pressure, and vice versa. That is to say, if a stream of perfect

liquid flows in a frictionless pipe of gently varying section, the
pressure increases as the size of the pipe increases and decreases
as the size of the pipe decreases. This is demonstrable in the case
of flow through pipes, although it is necessary to have the changes
of section very gradual in order to obtain the smooth continuous
motion to which alone the steady motion formula is applicable.
3. Application of Steady Motion Formula to Ships. The
steady motion formula applies to the motion of a liquid, including
motion past a solid at rest. In the case of ships, we are interested
in the motion of a solid through a liquid at rest. The two cases
are,however, as already stated, mathematically interchangeable.
Suppose we have a ship moving uniformly through still water
which extends indefinitely ahead and astern. If we suppose both
ship and water given the same velocity, equal and opposite to- the
velocity of the ship in the still water we have the ship at rest and
the water flowing past it. The mutual reactions between ship

and water are identical whether we have the ship moving through
stillwater or the water flowing past the fixed ship. To the latter
case, however, the steady motion formula applies if we neglect

friction and the mathematical treatment is much easier.

If the ship is in a restricted channel so shallow and narrow that

the area of the midship section of the ship is an appreciable fraction
of the area of the channel section, the steady motion formula
teaches us that with the water flowing past the fixed ship there
will be abreast the central portion of the ship where the channel

area diminished an appreciable increase in velocity of flow and


reduction of pressure.
The surface being free, reduction of pressure would result in

depression of surface. Passing to the case of the ship moving

through the channel we would infer that the water is flowing aft
abreast the central portion of the ship and that there is a depression
in this vicinity.

This, as a matter of fact, occurs in all cases, but in open water the
motions are not so pronounced, and it is seldom possible to detect
them by the eye. In a constricted channel, however, it is generally
easy to detect the depression abreast the ship since it extends to
the banks. If these are sloping the depression shows more plainly
than it does against vertical or steep banks.
There might be quoted many other illustrations of the validity
of the steady motion formula taken from phenomena of experience.
There no doubt of its general validity within certain limits as

regards motion of water around solids, but in considering any par-

ticular case it should not be applied regardless of its limitations.

4. Failure of Steady Motion Formula. The steady motion

formula assumes frictionless motion. Water is not frictionless,
but its friction is not sufficiently great in the majority of cases to
seriously affect steady motion directly.
The main failure of the steady motion formula as regards prac-
tical cases is in connection with the transformation of pressure into
velocity and viceversa. Neglecting variations of level the steady

motion formula is
= a constant. By the formula the
W 2g
greater the velocity the less the pressure, and if the velocity be

made must become negative. Now,

sufficiently great the pressure
negative pressure would be a tension, and liquids are physically in-
capable of standing a tension. Hence, when the case is such that the
steady motion formula would give a tension the motion that would
be given by the steady motion formula becomes impossible and
the formula fails. In practice, in such a case, instead of steady mo-
tion we have eddying, disturbed motion. In fact, in actual liquids,
when the motion is such as to cause a reduction of pressure, eddy-
ing generally makes its appearance some time before the pressure
becomes zero. But for moderate variations of pressure we find for
actual liquids pressure transformed into velocity according to the

steady motion formula with great accuracy. The transformation

of velocity into pressure, however, according to the steady motion

formula, without loss of energy, is not common in practice. For

instance, experiments at the United States Model Basin have
shown that air will pass through converging conical pipes with
practically no loss of head except that due to friction of the pipe
surface. But when passing through diverging cones, even when the
taper is but one-half inch of diameter per foot of length, there is
material loss of head beyond that due to friction. It appears
reasonable to suppose that the difficulties found in converting
velocity of actual fluids into pressure without loss of energy are
connected with the friction of the actual fluids, both their internal
friction or viscosity and their friction against the pipes or vessels

containing them.
To sum up, we appear warranted in concluding that in flowing
water pressure will be transformed into velocity according to the
steady motion formula with little or no loss of energy in most
cases, provided the pressure is not reduced to the neighborhood
of zero, and that velocity will be transformed into pressure but with
a loss of
energy dependent upon the conditions.
It evident that if the total head or average pressure is great,

given variations of pressure and velocity can take place with closer
approximation to the steady motion formula than if the total head
be small.
5. Sink and Source Motion. The mathematics of fluid motion
or hydrodynamics being somewhat complicated will not be gone

into here, but results will be given in a few of the simplest cases
which are of interest and have practical bearing. Suppose we
have liquid filling the space between two frictionless planes which
are very close together. The motion will be everywhere parallel
to the planes, and hence will be uniplanar or in two dimensions

only. Suppose now that liquid is being continually introduced

between the planes at some point. It will spread radially at an
equal rate in every direction. The point of introduction of the
liquid is called a "source." Fig. i indicates the motion, S being
the source. If liquid were being abstracted at S the motion at

every point would be directly opposite that shown in Fig. i and 5

would be what is called a "sink." The sink and source motion is
not physically possible because the steady motion formula applies,
and for velocity and pressure finite at a distance from 5 the velocity
at 5 would be infinite. Butbe seen presently that the mathe-
it will

matical concept of sinks and sources has a bearing upon possible

motions. Suppose that instead of a single source or sink we have
A and a sink of equal strength at B. Liquid
in Fig. 2 a source at
is being withdrawn at B at the same rate at which it is being intro-
duced at A and in time every particle introduced at A must find
its way out at B. The motion being steady the paths followed are
stream lines. These paths are arcs of circles. A number of these
circular arcs are indicated in Fig. 2. They are so chosen that the
" "
flow or quantity of fluid passing between each pair of circles

is the same. Adjacent to the line connecting the sink and source the
path is direct, the velocity great and the circles close together. As we

leave this line the path followed from source to sink is circuitous,
the velocity low and the spacing of the circles greater and greater.
6. Sink and Source Motion Combined with Uniform Stream. -

Suppose, now, that the liquid in which the source is found is not at
rest but is flowing with constant speed from right to left. Fig. 3
shows the result of the injection of a source into such a uniform
stream. In this case we have a curve of demarcation DDD
rating the liquid which comes rom the source and the other liquid.
No liquid crosses this curve. Now, the motion being frictionless
it makes no difference whether DDD is an imaginary line in the

moving liquid or the boundary of a frictionless solid. Hence if in


a uniform stream we put a frictionless solid of the shape the DDD

motion outside of it will be the same as in Fig. 3. This motion
will if we could have a frictionless solid like
be completely possible
DDD, since we no
longer have the source with its impossible con-
ditions as regards velocity and pressure.
In Fig. 3 DDD
extends to infinity. Suppose, now, in a uniform
stream we put a sink and a source of equal strength as at A and B
in Fig. 4. The direction of flow of the uniform stream is supposed
parallel to AB. In this case the closed oval curve CCC
the liquid which appears at the source and disappears at the sink
from the liquid of the uniform stream. Hence, if a frictionless
solid of theshape of CCC took the place of the liquid inside the
oval the motion of the stream outside would be unchanged.
Of course the shape and dimensions of CCC would vary with
the relative strengths of source and sink and velocity of stream.
Instead of one source and one sink we may distribute a number
along the line A B enabling us to modify the shape and proportions
of the line of demarcation CCC. The author (see Transactions of
the Institution of Naval Architects for 1894 and 1895) nas extended
this method to cover the case of an infinite number of infinitely
small sources and sinks, thus enabling us to determine lines of
demarcation or stream forms both in plane and solid motions,
closely resembling actual ships' lines. Not only the stream forms
but also the velocities and pressures along them can be determined,
but the process is laborious and has not so far been given sufficient

practical application to warrant following further here.

The closed ovals due to a source and a sink in a uniform stream
somewhat resemble ellipses as appears from Fig. 4.
7. Flow in Two Dimensions in Practice. While to reduce the
motion to one plane or two dimensions, the assumption was made
that it took place between two frictionless parallel planes so close

together that the space between them practically constituted a

single plane, it shouldbe pointed out that motion practically iden-
tical with plane motion occurs in practice. Suppose we have a
body of cylindrical type of infinite length moving in some direction

perpendicular to its axis. The motion past will be identical in all

planes perpendicular to the axis.


The motion past an actual body of cylindrical type whose length

though not infinite is great compared with its transverse dimensions

will, over a great portion of the length, be practically the same as
ifthe length were infinite. A propeller strut is a case in point.
Ideal plane flow has direct practical bearing upon the motion past
such fittings.
8.Stream Lines past Elliptic Cylinders. One general case of
uniplanar motion that has been solved mathematically is that of an
elliptic cylinder moving parallel to either axis in an infinite mass of

liquid. The circle is a special case and a plane lamina is another

special case where one axis of the ellipse is zero. The general
mathematical formulae expressing the motion of an elliptic cylinder
through liquid may be referred to in Lamb's Hydrodynamics,"
edition of 1906, Article 71. They do not give directly the stream
lines past an elliptic cylinder but the latter can be deduced from

them. Figs. 5 to 15 show plane stream lines or lines of flow past

various types of elliptic cylinders. The lines in the first quadrant
only are shown
as they are symmetrical in the other three. The
proportions of the ellipses are given, the semi-major axis being
always taken as unity. Fig. 10 shows flow around a circular
cylinder and Fig. 15 flow past a plane lamina of indefinite length
and unit half breadth. The flow around a lamina is, however,
impossible since the formula would require an infinite velocity
around the edges, or, as indicated in Fig. 15, the stream line spac-
ing in the immediate vicinity of the edge would become infinitely
9. Pressure Variations around Elliptic Cylinders. Figs. 16 and
17 give some idea of variation of pressure along the central stream
line and around the surface of the cylinders. A particle approach-
ing a cylinder along the axis steadily loses velocity and gains pressure
until it comes to rest against the cylinder when its pressure is in-
creased by the total velocity head of the undisturbed stream. The
particle then starts around the cylinder, rapidly gaining velocity
and losing pressure until at a point where it has moved but a short
distance around the cylinder it has regained the velocity and re-
turned to the pressure it had in the undisturbed stream. The
velocity then continues to increase and the pressure falls as shown

until the particle isabreast the center of the cylinder when the

velocity is at a maximum and the pressure at a minimum.

Figs. 16 and 17 show negative pressures but these are only

relatively negative. For convenience the diagrams are drawn as
if the pressure in the undisturbed stream were zero. The actual
pressure in any case is the pressure of the figure with the pressure
in the undisturbed stream added. Bearing in mind also that in
each figure the unit of pressure is the pressure head due to the
velocity of the undisturbed stream, or the velocity head of the
stream, Figs. 16 and 17 shed a good deal of light upon the effect of
variation of proportions. Thus, for an ellipse one-tenth as wide
as long, the maximum reduction of pressure abreast the center is
about one-fifth the velocity head. For the ellipse four-tenths as
wide as long, the maximum reduction is nearly the velocity head.
For the ellipse as wide as long (the circle), the reduction is three
times the velocity head. For the ellipse two and one-half times as
wide as long, the reduction is over eleven times the velocity head,
and for the ellipse five times as wide as long, the reduction is thirty-
five times the velocityhead and about one hundred and seventy-
five times the reduction for the ellipse one-tenth as wide as long.
The velocity head being proportional to the square of the speed,
the reduction in or increase of pressure at every point is propor-
tional to the square of the speed, and hence if any of the cylinders
were pushed to a high enough speed the reduction of pressure
abreast the center would equal the original pressure in the undis-
turbed stream, and hence the pressure abreast its center would
reduce to zero resulting in eddying. But eddying would appear
in the case of an actual cylinder long before the pressure abreast
the center became zero. For the excess velocity amidships would
not be fully converted into excess pressure on the rear of the cylinder
as required for perfect stream motion, and eddying would show
itself aft.

10.Disturbance Abreast Cylinder Centers. It is evident from

Figs. 1 6 and 17 that in the case of a cylinder moving through still

water the maximum sternward velocity of the water at any point
of the cylindrical surface is abreast the center of the cylinder. It
is also true that for motion parallel to the axis of x the greatest

sternward velocity for any value of y is on the axis of y. It is of

interest to trace the variation of velocity as we pass along the axis

of y. Fig. 18 shows sections of seven types of cylinders ranging
from the flat plate, No. i, which is all breadth, to the circle,
No. 4, and the ellipse five times as long as wide, No. 7. They all
have unit half breadth on the axis of y and are supposed to move
with velocity V parallel to the axis of x.
Fig. shows also curves of sternward velocity u of the water

as we pass out from the cylinder along the axis of y expressed as

a fraction of the speed of advance of the cylinder. It is seen that
the long cylinder causes the minimum disturbance at the surface
of the cylinder where y = i, but the maximum beyond y = 4.

Fig. 1 8 shows markedly the very great variations of disturbance

in the vicinity of the cylinder with variation of ratio of breadth to

length. The areas of the curves of Fig. 18 are the same, being

equal to V X (half breadth). The dotted square in the figure

shows this area.
u. Tracks of Particles. While Figs. 5 to 15 show stream lines
or flow past the cylinders, they give little idea of the paths followed

by particles of water when a cylinder is moved through water

initially at rest.
Rankine gave, many years ago, the differential equation to these
paths for the motion of a circular cylinder, and while this equation
cannot be integrated it is possible by graphic methods to determine
the resulting paths with ample accuracy.
Fig. 19 shows the paths followed by a few particles at various
distances from the axis as a cylinder of the size indicated by the
dotted semicircles in the figure passes along the axis from an infinite
distance to the right to an infinite distance to the left.
A on each path shows the original position of the particle when
the cylinder is at an infinite distance to the right. B, C, D, E
and F on the paths of the particles show positions when the cylinder
is at B, C, D, E, and F, on the axis as indicated.

The paths are symmetrical, and G denotes the position of each par-
ticle when the
cylinder has passed to an infinite distance to the left.
Fig. 19 shows the curious result that each particle is shifted
ultimately a certain distance parallel to the direction of motion of

the cylinder. This could not occur if the cylinder started from
rest at a finite distance from the particle, and came to rest within
a finite distance of the particle. For such motion the particles must
on the average be slightly displaced in a direction opposite to the
direction of motion of the cylinder.
12. Stream Lines around Sphere. While there are very few
mathematical determinations of stream lines in three dimensions
those for the sphere are known and it is of interest to compare them
with those for a circular cylinder shown in Fig. 10. The stream
lines pasta sphere are identical in all planes through the axis parallel
to the direction of undisturbed flow.

They shown in Fig. 20, and in Fig. 21 are shown curves of


pressure variation along the horizontal axis and around the sphere,
also along the horizontal axis and around a circular cylinder. The
curves show as might be expected that the sphere creates less dis-

turbance. This is evidently because the water is free to move in

three dimensions around the sphere, while it is restricted to plane
motion around the cylinder.
increase of pressure in front of the sphere is less. There is
a sudden rise close to the intersection of axis and sphere. At this
point the increase of pressure is the same as in the case of the
cylinder, being the pressure head due to the undisturbed velocity.
Abreast the center the loss of pressure is one and one-half times that
due to the velocity as contrasted with three times the velocity head
in the case of the cylinder. In other words, if a sphere is advancing
with perfect stream line action through water otherwise undisturbed
the water abreast the center is flowing aft with one-half the velocity
of advance of the sphere. In the case of the circular cylinder the
water abreast its center flows aft with velocity equal to the velocity
of advance.

2. Trochoidal Water Waves

Mathematical Waves.
i. Ocean waves during a storm are
generally confused rather than regular. They are not of uniform
height or length from crest to crest, and the crests and hollows
extend but comparatively short distances. After a storm, how-
ever, the confused motion settles down into rather uniform and

regular swells and the motion approaches that of mathematical

waves. For mathematical treatment it is necessary to assume
regularity of motion. We
may define a series of mathematical
waves as an infinite series of parallel infinitely long identically
similar undulations advancing at uniform speed in a direction

perpendicular to that of their crests and hollows. The constant

distance between successive crests is called the length of the waves
or the wave length, the distance between the level of the crest and
the level of the hollow is called the height of the wave, and the time
interval between the passage of successive crests by a fixed point
is called the period of the wave.
Mathematical waves are cases of motion in two dimensions,
since the motion is identical in all planes perpendicular to the
wave crests.

2. Trochoidal Wave Theory. The most commonly accepted

theory of regular wave motion is that called the trochoidal

theory." Its too long and difficult to be gone

mathematics is

into here, and I shall undertake only to give some of the formulae
and conclusions that have been evolved by the eminent mathe-
maticians who have worked in this field.
Of the British mathematicians who have contributed to the
trochoidal theory, Airy and Rankine were especially prominent

shortly after the middle of the last century.

By the trochoidal theory, in water of unlimited depth each
particle describes at a uniform rate a circular orbit, making one
complete revolution per wave period, the radii of the orbits being
a maximum for surface particles and decreasing indefinitely with
Referring to Fig. 22 let the wave length be denoted by L and
let R be the radius of a circle whose circumference is L. Then
R= -
Suppose we locate this circle with its center midway
2 7T

between the levels of crest and hollow and take a point P on the
radius at a distance r or from the center, H being the wave

height. Then, if the circle rolls on the line AB the point P will
describe a trochoid giving the outline of the wave surface. This

trochoid shows the contour assumed by particles originally at the

surface level.Similarly, particles originally at any level below the
surface are found along a trochoidal surface having the same
diameter of rolling circle but less orbit radius, the radius diminish-

ing indefinitely with depth.

Fig. 23 shows the trochoids at various levels, orbit diameters
and contours of lines of particles which in undisturbed water were
equally spaced verticals. The cycloid the limiting trochoid
is shown, but not possible for sharp crested waves to appear
it is

in practice. They break long before they approach closely the

limiting cycloid.
is for water of unlimited depth.
Fig. 23 In water of finite depth,
by the trochoidal theory each particle describes an elliptical orbit
instead of the circular orbit of deep water. Referring to Fig. 24
let A BCD be the rolling circle" whose perimeter, as before, is
equal to the wave length from crest to crest. Let the ellipse
EFGH of center the same as the center of the rolling circle be the
orbit of the surface particles. Let OP' be the radius of a concentric
circle of diameter the same as the major (horizontal) axis of the
ellipse. Then, as the rolling circle moves, let the radius OP'
revolve with it and the ellipse move horizontally with it without
revolving. Draw vertical lines as P'N from the successive posi-
tions of P' to meet the ellipse in points such as P. The modified
trochoid obtained by joining all points such as P is the surface profile
of the wave.
The horizontal and vertical axes of the elliptical orbits are not
independent but vary with the depth of water, the depth below
the surface, etc.
Thus let a and b denote the horizontal and vertical semi-axes,
respectively, of an elliptical orbit whose center is a distance h
below the orbit centers Let oo&o denote
of the surface particles.
the semi-axes of the surface orbit. Let d denote the depth from
center of surface orbits to the bottom. Let R denote the radius
of the rolling circle and w the angular velocity with which it must
roll to have its center travel at the speed of the wave.
Let L denote the wave length in feet, v the wave speed in feet

per second, g the acceleration of gravity and e the base of hyper-


bolic logarithms. Then the formulae connecting the above quan-

tities are as follows:

e L i

e L e
b = -2?rrf

g L e L

1x(d-h) -2*(
L +e
e L e L

and w = m/
a R

and if T denote the period in seconds

bo g

To pass to the case of indefinitely deep water, we put d

= oo .

Then aQ = b = r say, and if r denote the radius of the circular


orbit at a distance h below the surface orbits, we have

a = b = r = r e L .

As before, v
= &R, but

R 2 TT

Substituting for g the value 32.16 and for TT its value, we have the
following formulas for deep-water trochoidal waves:

Velocity in feet per second

= v = 2.26
Velocity in knots
= V = 1.34
Period in seconds = T= 0.442
Length in feet 5.118 r = .557 V 2 = 2

The above rather complicated-looking formulae express com-

pletely the motion under the trochoidal theory.

3. Mechanical Possibility of Trochoidal Waves. For the

motion to be possible it must satisfy,
The condition of continuity.

The condition of dynamical equilibrium.


3. The boundary conditions.

4. The conditions of formation.

The mathematical investigation of the above conditions is too
long and complicated to be given here. The results only can be
given. As regards continuity, it is found that the motion is possible

in water of infinite depth, but that in water of finite depth the

equation of continuity is not quite satisfied.

As regards dynamical equilibrium, again we find that the motion

is not quite possible in finite depth, the pressure at the surface being

not quite constant, which it must be from boundary conditions*

In however, the pressure as deduced from the tro-
infinite depth,

choidal formulae is constant along the wave profile and hence the
motion is possible.
The only other boundary conditions to be satisfied are those at
the bottom, and these are satisfied by the trochoidal formulae,
since they give at the bottom horizontal motion only (b = o) when
the water is of finite depth and no motion at all (r = o) when the
water is of infinite depth.

Finally, as regards the condition of formation, it is a theorem

ofhydrodynamics that a perfect liquid, originally at rest, that has
been acted upon by natural forces only, cannot show molecular
rotation. The trochoidal wave motion involves a slight molecular
rotation, and hence falls slightly short of being a possible motion
in both finite and infinite depths.
We conclude, then, that trochoidal wave motion falls slightly
short of being mathematically possible; but it would require a very
small change in the motion to render it possible. This and other
considerations which will be pointed out later warrant the adoption
of the trochoidal theory as a working approximation.

4. Trochoidal Wave Profiles. The formulae already given may

be supplemented by those representing the trochoidal contours at
various depths. They are x = Rd a sin 6, h y =
b cos 6, where
x is measured horizontally, y is measured vertically down from the

surface orbit centers R, a and b have the values already given and
6 is angle rolled through by the rolling circle, being = o for an
initial condition where the radius of the rolling circle is vertical and
its center ;under the crest of the trochoid. Of course, in deep water
a = b = r.

Fig. 25 shows the wave surface profiles for three waves, each 300
feet long and 20 feet high, but in three depths of water, namely oo ,

25 feet and 15 feet. These three profiles have the same line of
undisturbed water level. It is seen that in each case the orbit
center, or mid height of wave, is above the level of the undisturbed
water. For deep-water waves the amount of this elevation is
- For shallow- water waves
> r being the surface orbit radius. it
2 R
The pressure on any trochoidal subsurface for deep-water

waves is uniform and the same as the pressure in undisturbed water

on the corresponding layer.
For subsurface trochoids the elevation of orbit centers is given
f(? R
by e ,
where h is the distance of the orbit centers from the

level of surface orbit centers.

5. Energy of Trochoidal Waves. Consider now the energy of

waves in deep water. This is partly potential, due to the fact that
in wave motion the particles are elevated on the average above their
still-water positions, and partly kinetic, due to the velocity with
which the particles of water are revolving in their circular orbits.
Let w denote the weight of one cubic foot of water. Then the
potential energy of a mass of water one foot wide and one wave
length long, i.e., extending from one crest to the next, is

where r is surface orbit radius or one-half the wave height.

Now R = - -
Substituting this value we may write
2 7T

\ L~

In practice, for actual waves is a small fraction and for


most purposes can be ignored. The kinetic energy of the mass of

water as above is exactly the same as the potential energy, or if
we denote it by Ek ,

v (

While the potential and kinetic energies of a mass of water in

wave motion remain constant, there is constant transmission of
energy going on.
Fig. 26 shows a number of positions of a distorted vertical or
line of particles originally vertical in still water.
During part of
the motion, energy is being transmitted across this vertical in the
direction in which the wave is traveling and during the rest of the
motion it is being transmitted backward. One wave length away
is a similar distorted vertical moving in the same way, so there is

at no time net gain or loss of energy to a mass of water one wave

length long. But the energy transmitted forward across a surface

originally a vertical plane is during one wave passage greater than

the energy transmitted backward -

by the quantity ( i -

4 \ J-i /J

This is identical with the kinetic or potential energy of the wave,

so that a mass of water extending over one wave length receives
from the water behind it and communicates to the water in front
of it during the passage of one wave a net amount of energy equal
to its kinetic or potential energy.
While this is the netenergy transmitted the rate of transmission
is much higher during a portion of the wave passage than the aver-
age. Thus, if 6 is the angle in its orbit from the vertical of the
radius r of a surface particle, the rate of transmission of energy

through the distorted vertical terminating in the surface particle

(see Fig. 26) is given by

By integrating this between the limits 6

= o and 2 IT, we get the
expression given above for the net energy transmitted. Fig. 27

shows a curve of rate of transmission of energy for a deep-water

wave 300 feet long and 20 feet high. Between o and 90 and
270 and 360 there positive transmission.
is Between 90 and
270 there is negative transmission. The average rate of trans-
mission is indicated on the figure.
6. Superposition of Trochoidal Waves. If we superpose two

trochoidal wave same length L, and hence the same

series of the

speed of advance, which are traveling in the same direction with

parallel crests a distance a apart, the result is a single series of
length L.
If we denote by HI, HZ the wave heights of the two components

and by H the height of the resultant series, we have

Evidently if a = o, or the crests of the component series are im-

mediately over one another, cos =i and H 2

=(Hi+ H^) 2 . In
this case the wave height of the resultant series is the sum of

the component heights. If a = irR we have cos = i and

H z = Hz)
In this case the crest of one component is

immediately over the hollow of the other, and the height of the
resultant series is the difference of the heights of the components.
If in this case Hi=H z the components extinguish each other and

the resultant is still water.

7. Wave Groups. A very important deduction from the tro-
choidal theory is the theory of wave groups. If we superpose
two trochoidal systems of equal heights, but slightly different
lengths, we have at one point of the resultant series waves of
double the height of either component and at another point waves
of zero height, since at one point of the series we would have crest
superposed on crest and at another point crest superposed on
hollow. The resultant series in this case would consist of a number
of groups of waves, each with a wave
height in the of maximum
middle and of heights steadily decreasing ahead and astern of the
middle until waves of infinitesimal height or bands of practically
still water separate the groups. It can be easily proved from the

trochoidal theory that each group will travel as a whole at just half
the speed appropriate to the wave length of the original compo-
nents. The individual waves, however, travel at their natural

speed, which is double the group speed. A

wave will advance
from the rear of a group where its height is infinitesimal and pass
through the group, growing until it reaches a maximum at the
center of the group and then dwindling as it goes forward until
its height again becomes infinitesimal at the front of the group.

One can readily start a group of circular waves by dropping a

pebble from a bridge into a placid stream. This shows general
features somewhat similar to the theoretical trochoidal group.
If the reflection in the water of the side of the bridge is distinct a

wave can be watched as, first becoming noticeable at the rear, it

passes through the group, reaching a maximum height and dying

down again. as it gets further and further ahead of the center of
the group. It will be found, however, that unlike the theoretical
trochoidal group, which has similar groups some distance ahead
and astern of it, the circular group gets wider and wider from front
to rear. If, for instance, at a given time it shows five appreci-
able waves, it will be seen a little later to show six, then seven, and
so on.
8. Applicability of Trochoidal Theory. Having considered the
nature of the motions and the conclusions that can be drawn from
the trochoidal wave theory, it is time to consider its applicability
to actual water waves. We know that actual waves cannot be
exactly trochoidal, and we are not warranted in assuming without
some confirmatory evidence that the trochoidal theory gives us
waves substantially the same as actual waves. Now, as already
pointed out, actual waves are almost never regular, so that a rather
rough approximation, mathematically, to the ideal regular waves
would, as a rule, resemble them more closely than do the actual
waves. Hence, if we find that the trochoidal theory adequately

represents the most important feature or features of wave motion

we need not be concerned as to minor features.
Stokes has developed a mechanically possible theory of wave
motion where the wave profiles are sines and the speed of the wave
is not independent of the height, but increases slightly with it.

For waves of ordinary proportions, however, the speed is practi-

cally the same as by the less complex trochoidal theory.
It appears, then, that for the proportions occurring in practice
trochoidal waves are in substantial agreement with mathematical
waves free from minor mechanical imperfections.

Now, what is the basic feature of trochoidal waves? It seems

that it may fairly be said to be the fact that the velocity of advance
depends only upon the length from crest to crest and the depth of
the water. We have seen that the formula for this velocity is

2nd -2xd

2 7T ^+
e L e
2 7T

Small-scale experiments in tanks, such as those of the Weber

Brothers, who published their results in 1825, have given results
consistent with the trochoidal theory; but it is obviously desirable
to compare the theory with actual full-sized waves, which it is

very do with accuracy.

difficult to

9. Gaillard's Experimental Investigations of Trochoidal Theory.

Major D. D. Gaillard, U. S. A., in a monograph on Wave

Action in Relation to Engineering Structures (Professional Papers,
No. 31, Corps of Engineers, U. S. Army), has compared reported
speeds of advance and speeds computed by the trochoidal theory
in eighty-five cases of ocean waves observed by various people at
various places. Of these eighty-five reported velocities, twenty-
three were higher than the computed velocities corresponding to
the observed length and sixty-two were lower, the average of the
whole number being nearly 9 per cent below the average computed
velocity. While giving due consideration to the difficulties in the
way of accurate observation, the agreement between these observa-
tions and the trochoidal theory is certainly not wholly satisfactory.
Fortunately, Major Gaillard gives a further comparison of the
trochoidal theory with a large number of observations, taken by
himself or under his direction, under conditions favorable to

accuracy. These observations were made in 1901 and 1902 in the

Duluth, Minn., ship canal and in Lake Superior near the canal.
The canal in question is about 300 feet wide, 26 feet deep,

where the observations were taken, and about 1000 feet long. It
connects the harbor of Duluth with Lake Superior, and natural
conditions are such that during and after storms, waves often

pass squarely into its mouth and on through it. By means of

instantaneous photography accurate profiles of waves against the
walls, either in the canal or outside, in gently shoaling water,
could be determined. The velocity of the waves could also be
determined quite accurately, velocity observations being usually
taken between stations 300 feet apart. The observations during
two years numbered 631 in all. The wave heights varied from 2
to 23 feet, the wave lengths from 45 to 425 feet, and the wave
velocities from 9.1 to 33.3 feet per second. The depth of the
water varied from 3.3 to 27 feet, though 533 of the observations
were taken in the canal 26 feet deep. For these 533 observations
the mean observed velocity and the mean velocity as computed
from the shallow-water trochoidal formula agreed within less than
one-half of one per cent. This is practically exact agreement.
For the ninety-eight observations made outside the canal in varying
depths the computed velocities averaged nearly 5 per cent more
than the observed velocities. Major Gaillard states that conditions
and facilities were such that the last series of observations could
not be taken with the same degree of accuracy as those on waves
inside the canal. Major Gaillard's observations appear to furnish
conclusive evidence of the reliability of the trochoidal theory as
regards its most important feature, the relation between length
and speed of advance.
It is true that Major Gaillard dealt only with shallow-water
waves, but it is evident from what has gone before that shallow-
water trochoidal waves are more likely to misrepresent the actual
waves than the deep-water trochoidal waves.
The actual wave profiles in the Duluth canal as obtained by
photography agreed reasonably well with the profiles from the
trochoidal formula. The differences, speaking, were
greatest at about mid-height of the wave, where the failure of the
trochoidal theory to satisfy the conditions of continuity and

dynamical stability is most marked. Major Gaillard states that

the elevated portion of an actual wave " is always narrower and

the depressed portion broader and flatter than is indicated by

theory, and this difference becomes more marked as the wave
approaches the point of breaking." The actual wave profiles,
however, were by no means uniform, differing from each other
quite as much as from the trochoidal form.
To sum up it seems fair to say that the trochoidal formulae
represent actual waves very closely as regards speed, with a suffi-

and for practical purposes

cient approximation as regards profile,
are much better than more complicated and difficult formulas that
have been devised. They are themselves quite complicated and
Shallow Water and Solitary Waves.
10. The trochoidal for-
mula for wave speed in shallow water of depth d may be written

L - i
gL '

For a constant length of wave v decreases as the water shoals, the

ratio between the velocity of a wave of given length L in water
of depth d below orbit centers and a wave of the 'same length in

indefinitely deep water being

Fig. 28 shows a curve of the value of this ratio plotted on It


is seen that for depths of water greater than half the wave length
there is no change of speed.
Figs. 29 and 30 show graphically the relations between depth of
water, length of wave and speed of wave, the speeds being ex-
pressed in knots per hour. Fig. 30 simply reproduces on a large
scale for clearness the lower part of Fig. 29. It is seen that as the

depth of water becomes very small the speed tends to become

independent of the length. So let us investigate the results of
assuming that the wave length is very much greater than the depth
of water.

The formula for wave speed in shallow water is, as we have seen,

do 2 7T
i^ 2 7T
L + I

Now expanding we have


2 7T
f +

Now when > or the ratio between depth and length, becomes very

small all terms of the long fraction above except two can be neg-
lected, and the fraction reduces to

Z= d
Then 1?= 2 w- * = gd.
L 2 7T

In the above not the original depth of water but the depth
6? is

to surface orbit centers, or to mid-height of the waves. This

depth is somewhat greater than undisturbed still water depth, but
not very much greater.
The above result is interesting as indicating that in shallow water,
on the trochoidal theory, there is a limit to the speed of waves no
matter what their length. This conclusion is confirmed by ex-

perience, and the value of the limit obtained above is in reasonable

agreement with experiments. It is interesting to note in this

connection that, as indicated in Fig. 25, the shoaler the water the
more a trochoidal wave system tends to approach a series of sharp
crests separated by long hollows that are nearly flat. That is to
say, it tends to become a series of solitary waves, or waves of
translation, consisting of humps or crests without hollows. Scott

Russell, as a result of numerous experiments on the so-called solitary

wave, or wave of translation, made in a trough, concluded that the

velocity of this wave was equal to that of a body falling freely
through a height equal to half the depth from the top of the wave.
The formula above gives the velocity of the trochoidal wave ap-
proaching the wave of translation type as that of a body falling
through a height equal to half the depth measured from mid-height
of the wave. The difference is not great for possible waves whose
height generally but a fraction of the depth.
is There is, however,
testimony to indicate that Scott Russell's formula gives too great a
velocity. Rankine gives a formula practically equivalent to Scott
Russell's.Major Gaillard states that he has applied Rankine's for-
mula to several hundred observations upon shallow-water waves,
taken at North Beach, Fla., and on Lake Superior, and has found
that it almost invariably gives results considerably in excess of the
observed velocities. The trochoidal formula, then, with its velocity
somewhat smaller than Scott Russell's or Rankine's, would agree
more closely with Gaillard's observations.
ii. Dimensions of Sea Waves. It may be well to supplement
the mathematical theory of waves with some information regard-

ing waves found in practice. The heights of sea waves are their
most striking feature and the most important for seagoing people.
From the nature of the case it is very difficult to observe with
accuracy the heights of deep-sea waves. From observations made
by a number of observers of various nationalities in various seas
it seems reasonable to consider that waves 40 feet high from
trough to crest can be generated in deep water by unusually severe
and long continued storms. This exceptional height is liable to
be materially surpassed by abnormal waves, the result of super-
position. Thus Major Gaillard quotes a case where a photograph
taken on the United States Fish Commission steamer Albatross,
and furnished him by Commander Tanner, U. S. N., showed the
fore yard of the ship parallel to the crest of a huge wave and a
little below it, the photograph being taken from aft. From the
known dimensions of the vessel and position of the camera it seems

that this crest must have been from 55 to 60 feet above its trough.
This wave was photographed in the North Pacific off the United

States coast. Estimated heights as great as this are not infre-

quently reported by captains of steamers crossing the Atlantic,
but accurate estimates of wave heights are difficult to make.
Probably it would be a fair statement of the case to say that very
heavy seas with maximum wave heights of 30 feet are not unusual.
Exceptionally heavy seas with maximum wave heights of 40 feet
is good evidence that abnormal
are encountered at times, and there
crests60 feet in height have been encountered. The maximum
wave height would not be found for every wave of a heavy sea.
The 30 and 40 foot waves would appear at intervals. Intervening
waves would be lower.
For the purpose of estimating the maximum stress of a ship
it is customary to assume a wave height one-twentieth the length,

the length of wave being taken the same as the length of the ship.
This seems a reasonable average, but steeper waves have been
often observed. Short waves are more apt to be steep than long
waves. As to actual lengths may be
confidently stated that

waves over 500 feet long are unusual, though a 4O-foot sea would
probably be between 600 and 800 feet long, and lengths of 1000
feetand more have been measured.
For the development of maximum waves
a great space of open
water is essential. Major Gaillard concluded after investigation
that "during unusually severe storms upon Lake Superior, which
occur only at intervals of several years, waves may be encountered
in deep water of a height of from 20 to 25 feet and a length of

275 to 325 feet." It appears, then, that the 5oo-foot vessels navi-
gating Lake Superior will probably never encounter waves their
own length. This condition indeed is rapidly being reached by
the enormously long Atlantic liners of the present day.
12. Relations between Wind and Waves. The length of waves
(or their speed of advance) is governed by the velocity of the wind
creating the wave. The relationis not known. Waves have
often been observed in advance of a storm and also waves in a
storm that were traveling faster than the wind was blowing. It
does not follow that a wave can travel faster than the wind that
forms it. Severe storms are revolving or cyclonic, and the storm
center does not move as fast as the wind blows. Hence a wave,.

though traveling more slowly than the wind that formed it, may
run entirely ahead of the storm or into a region where the wind is
blowing less violently.
Published observations upon the ratio of wave and wind velocity
are not very concordant. Lieutenant Paris, of the French Navy,
maker of very extensive and careful wave observations, gives the
wave velocity as .6 that of the wind in a very heavy sea, and
relatively greater as the sea becomes less heavy. Major Gaillard
found at Duluth for waves in shallow water, which probably did
not travel so fast as in the open lake, that the wave velocity as
averaged from observations taken during fourteen storms was but
.5 that of the wind. It appears probable that in a strong gale

making a heavy sea the wave velocity is from .5 to .6 that of the

wind, but that waves formed under these conditions often travel
to regions where the wind is not blowing so fast as the waves are

If we take the wave formed as moving with .5 the speed of
the wind we have from the trochoidal formula for deep water the
following relations:

Speed of wind, statute miles


From this formula we have the following :

h = 10 15 20 25 30 35 40
/ = 44 zoo 178 278 400 544 711
At first might appear inconsistent with the
sight these results
fact that waves more than 40 feet high are very rare, even where
there are several thousand miles of open water. As a matter of

fact, however, violent gales are revolving storms, and the violent
part of such storms is seldom more than five or six hundred miles
in diameter, so that Stevenson's formula is consistent with the

general facts.

3. The Law of Comparison

1. Principle of Similitude. Modern ideas of the resistance of

ships are based largely upon the Law of Comparison, or Froude's

Law, as it is generally called, connecting the resistance of similar
vessels. By judicious application of this law we are enabled to
determine, with fair accuracy, the resistance of a full-sized ship

from the experimentally determined resistance of a small model of

the same.
Froude's Law is a particular case of the general law of mechan-
ical similitude, definingthe necessary and sufficient conditions that
two systems or aggregations of particles that are initially geometri-
cally similar should continue to be at corresponding times not only
geometrically but mechanically similar. The principle of simili-
tude was first enunciated by Newton, but the demonstration now
generally acceptedwe owe to French mathematicians of the last
century. Mr. William Froude appears, however, to have developed
independently the particular form used to compare ships and
models and to have been the first to use the Law of Comparison
to obtain useful practical results.
2. Deduction of Law of Comparison. Suppose we have a
particle of a system whose coordinates referred to rectangular
axes are x, y and z. Let m denote the mass of the particle. If the

particle is moving, it will have at time t an acceleration parallel

to the axis of x, an acceleration -r* parallel to the axis of y and

similarly parallel to the axis of z. Let the components parallel
to x, y and z of the external moving force upon the particle be
denoted by X, Y and Z. Denote by 8x, dy and dz the resolved
motions parallel to the axes due to a small motion of the particle
along its path.
Then using the well-known principle of Virtual Velocities, the
differential equation giving the motion of the particle is

Suppose, now, we have in a second system, mechanically similar,

a corresponding particle of mass m' whose coordinates at time /',
corresponding to time / in the first system, are #', y', z' and whose
impressed force components are X', Y', Z', Its equation of motion
will be


If the motions of these two particles are geometrically and

mechanically similar, the equations of motion must be the same,

differing only by a constant factor. Now, for similar geometrical
motions we have a constant ratio between x and x', etc.

Suppose x' = \x, y'

= \y, z' = Xz.

Then d2x' = \d2x and so on.

Let m' = /j.m, p being the constant ratio of masses of the two
Let the corresponding times be in the ratio T or t' = Tt and

Substituting for x , etc., their values we have

*/ A (/ *V

This may be rewritten


Evidently, in order that this may become identical with the equa-
tion for the first system, we must -have

x -
* X T*
and similarly
r = M = z^

Y T 2
It follows, then, that the external forces on corresponding par-
ticles must bear a constant ratio to each other. Let F denote this
ratio. Then the necessary and sufficient relation for geometrical

and mechanical similitude of motion of the two particles is F = ^-

The same relation connects every corresponding particle of the
two systems, and hence the systems as a whole. Now T, the
relation ratio between corresponding times, is not very convenient
for use in practical application. It is readily eliminated. Let
v and v be corresponding velocities. Then

_ dx _ dx' X dx
~ f
~ ~ '

dt' dt' T dt
v X X2
Whence = = c say. Then Tz =
1) JL C

Whence F = ^\
= ^-

We may further simplify the case by assuming a relation between

c and X. Suppose we make the ratio of corresponding speeds such
that c z = X or that the speed ratio is equal to the square root of
the dimension ratio. Then F = /*. Now we know that whatever
the speed ratio and dimension ratio, the external forces due to
gravity must be in the ratio /* or the ratio of masses. We see from
the above that for motions mechanically and geometrically similar,
ifthe speed ratio is made equal to the square root of the dimension
ratio, all external forces must be in the ratio of mass or weight.
The application to the case of a ship and its model is obvious.
If a certain portion of the resistance of a ship is due to a certain

disturbance of the water and if, at a corresponding speed of the


model, bearing to the speed of the ship a ratio equal to the square
root of the dimension ratio between model and ship, there is a
similar disturbance set up by the model, the resistances due to the
similar disturbances will be proportional to the weights of ship and
For the resistance of the ship or model, as the case may be, is in
each case the external force, other than gravity, acting upon the
system of particles involved in the disturbance, and the mass of
disturbed water, if the disturbances are similar, is proportional to
the displacement of the ship.
It is apparent from the above that the applicability of Froude's
Law to resistances of model and ship depends upon whether the
disturbances at corresponding speeds are similar. This is a matter
capable of reasonably close experimental determination as regards
the wave disturbances of model and ship. It is found that these
are similar at corresponding speeds, the wave disturbance set up

by the ship being an enlargement to scale as closely as can be

measured of that of the model at corresponding speed.
Mr. William Froude estimated the actual resistance
of the Grey-

hound, a ship of over 1,000 tons displacement, by applying the

Law of Comparison to carefully measured resistances of a small
model in a manner to be explained later, and found the results thus
obtained in very close agreement with the actual resistance as
measured by towing experiments. But, perhaps, the strongest
experimental confirmation of the Law of Comparison, and one fully
warranting its
practical application, is an indirect one. There are
now a number of experimental model basins in existence engaged
in estimating the resistances of ships by proper application of the
Law of Comparison to results of model experiments. These are
not able to verify their results directly, because, for the full-sized
ship when tried, we ascertain not resistance but the indicated power.
The efficiency of propulsion connects the indicated power with
the resistance. But, using the actual indicated powers and the
estimated resistances determined from model results by the Law
of Comparison, there are obtained efficiencies of propulsion which
new designs of vessels.
are consistent and reliable as a basis for
We are fully warranted, then, by numerous considerations,

both theoretical and practical, in reposing especial trust and con-

fidence in Froude's Law. The modern theory of ships' resistance
is founded upon it, and since it has been understood and utilized

the numerous crude and treacherous theories which preceded

Froude have practically disappeared.
It is possible to make a less general demonstration than the above
of Froude's Law from the steady motion formula for stream lines.

This, too, depends upon the similarity of stream lines around model
and ship, a fact requiring experimental determination.
3. Applications of Law of Comparison. Let us now determine
the formulae, etc., needed in the application of the Law of Compari-
son to ships' resistance.
Put into symbols, let L, B, H denote the length, breadth and
mean draft of a ship in feet, D its displacement in tons and V its
speed in knots. Let /, b, h, d, v denote similar quantities for a
model of the ship. Suppose R and r denote resistances following
Froude's Law. If X denote the ratio between linear dimensions
so that L = \l, B = \b and so on and if V and v are connected by
the relation V= v V\,
R r

Since T= T = x
a \//

we may write R = /L\ r = XV. It is to be noted, too, that

\ yI

\=(-\ so V = v(^

The Law Comparison is useful and applicable in connection

with many problems besides that of the resistance of ships. Thus,
it is directly applicable in comparing full-sized machines and their

models of the same material. Here, too, since gravity is one

external force always present, the speeds of corresponding parts
must be in the ratio of the square roots of the linear dimensions.
Thus consider a small and alarge steam engine, similar and working
at corresponding speeds. Let us find from the Law of Comparison
the relations connecting pressures, revolutions, etc. Let R, T,

/, S and P
denote, respectively, revolutions per minute, torque,
indicated power, piston speed, and steam pressure for the large

engine, and r, t, i, s, p the same quantities for the similar small

engine or model. Let X denote the ratio of linear dimensions.
Then since the speeds must correspond, we have S = s V\.
Now S = stroke of large engine X 2 R,
s = stroke of small engine X 2 r.
Also stroke of large engine = X stroke of small engine. Whence

= X 7? But 9 also = VA. Whence R = r =
s r s VX
The total steam pressures on the pistons being the external forces
must be in proportion to X 3 and the piston areas are proportional
to X
.Hence P = \p. The
indicated horse-power is proportional
to the piston area, varying as X the steam pressure varying as X ,

and the piston speed varying as VA. Hence on combining these

three factors we have I = *X
3 5

. Now 7 is proportional to TR.

Hence the torque is directly proportional to the indicated power
3 5
varying as X and'

inversely proportional to the revolutions varying

Hence T=
The above relations apply directly to centrifugal fans. For
steam pressure we substitute the pressure at which the air is de-
livered. Also the quantity of air delivered will vary directly as
the area of outlet pipe or as X2 and directly as the speed or velocity
or as X*, whence at corresponding speeds the quantities of air
delivered will vary as X 2

The above relations for revolutions, torque, power and pressure

apply too to the operation of propellers. It should be noted since
P = \p that the pressure per square inch of the water in which a
propeller works should be X times that of the water in which its
model works. Model propellers are usually tested under a total
head of 35 feet or so of water (equivalent to atmospheric pressure -+-
one foot or so submersion below surface, say, 35 feet in all). For
the pressure to vary linearly would require a full-sized propeller

ten times as large as the model to work under a total head of 350
feet, or, say, 316 feet submersion, if the 34 feet head due to air

pressure were equivalent inall respects to 34 feet of water. While

this is
only approximately the case, it is evident that the pressure
conditions for model and propeller are not those required by the
Law of Comparison. But it does not necessarily follow that the
Law of Comparison would not apply to the conditions of practical
operation. the action of propellers is such that the power,

torque and efficiency are unaffected by depth of submersion, the

Law Comparison would apply fully.
We under some conditions of operation,
shall see later that,

propeller action is but little affected by depth of submersion,

while under others it is materially affected. Hence under some
conditions the Law of Comparison applied to model propeller

experiments may be expected to be a reliable guide, while under

other conditions of operation it would certainly be fallacious.
Valuable and even indispensable as the Law of Comparison is
in dealing with resistance and propulsion of ships, it must be

applied with discretion and an understanding of its limitations.

Some of these limitations will be developed later.

4. Simple Resistances Following Law of Comparison. In

reducing any kind of resistance to rule the endeavor is usu-
ally made to express it by a formula involving some power of the
speed as V or F"
2 3
Unfortunately actual resistances of ships do

not lend themselves to such simple formulae, but it seems worth

while to determine how resistances which satisfy the Law of Com-
parison and vary as definite powers of speed vary with displace-
ment or dimensions.
Suppose R = </>(Z)) Vn
expresses the law of variation of a ship
resistance which satisfies the Law of Comparison, R
being resistance
in pounds, <f>(D} some function V the speed in
of displacement,
knots and n an index according to which resistance varies.
For the similar models' resistance we have

r = <}>(<) v

For corresponding speeds ,

V= /Z>Y
I and
v \d/ r d


Lfi d6
Whence ~^r
= $W Td = a constant regardless of displacement
= C say.
Then <j>(D)=CD ~^
or R = CVnD ~\
For integral values of n we have the following results

n = i resistance varies as (displacement) 1 or (linear dimensions) 2 *.

n = 2 resistance varies as (displacement)* or (linear dimensions) 2 .

n = 1
3 resistance varies as (displacement)* or (linear dimensions) *.
n = 4 resistance varies as (displacement)* or (linear dimensions)

n = 5 resistance varies as (displacement)* or (linear dimensions)*.

n = 6 resistance is independent of displacement or dimensions.

The above results are not of much practical value since actual
resistances even when following the Law of Comparison do not vary
as simple powers of the speed, but they are of some use in connec-
tion with approximate formulae.

5. Dimensional Formulae. In connection with the Law of

Comparison it is of interest to note the so-called dimensional for-

mulae which are the functions of certain primary variables or
units to which are proportional a number of things which we shall
have occasion to use. Thus taking length or a linear dimension
as a primary variable we have area varying for similar surfaces
as (linear dimensions) and similarly volume varies as (linear
Then if we denote length or linear dimension
dimensions) .
by /
we have /
and /
as the dimensional formulae for area and volume

Similarly, denote time, since velocity varies directly as the
if /

length traversed in a given time and inversely as the time required

to traverse a given length and is dependent upon no other variables,

we have -
as the dimensional formula for velocity. Further, since

acceleration varies inversely as the time required to gain velocity we

have - as the dimensional formula for acceleration.


The application of dimensional formulse is mostly

in connection with conversion factors for the determination of
the numerical magnitude or numbers representing definite things
when the fundamental units are changed. Thus, suppose we have
a length of 24 feet. If the yard were the unit of length this length
would be expressed numerically by 8 instead of 24. Similarly,
suppose we have a surface of 108 square feet. If the yard were the

primary unit the number of units of surface would be -r-r-.2 = 12.


Since the dimensional factor for area is I? the conversion factor

is the square of the ratio of the linear units. Similarly the con-
version factor for volume is the cube of the ratio of linear units and

135 cubic feet would be -r

= 5 cubic yards. These transforma-

tions are puzzling in some cases and it will be well to give the
general rule applicable.
Wewill have in any given case the old number, or the number
expressing something quantitatively in the old units, the ratios
between the units or the numbers expressing the new units in
the old units and vice versa, and the dimensional formula for the
thing under consideration area, volume, velocity or what not.
Thenexpress the old unit of each kind in terms of the new and

the corresponding numerical ratio

version factor, and we have

substitute in the dimensional formula for each primary variable

new unit

The result is the con-

New number = Old number X Conversion factor.

Thus when converting square feet to square yards the ratio

en ^
length unit
- r = -
The dimensional formula is /
. Then

Conversion factor = (-] = -


Old number = 108.

New number = 108 X-= 12.


Similarly, suppose we have a velocity of 69.3 feet per second and

wish to convert it into statute miles per hour.

For velocity the dimensional formula is

Old length unit i Old time unit i

New length unit 5280 New time unit 3600

Conversion factor
= i
--- = -3600
-- =
i is
5280 3600 5280 22

New number = 69.3 X = 47.25 statute miles per hour.


By following the above method strictly and systematically there

is no difficulty in obtaining correct conversion factors no matter
how complicated the dimensional formulae.
It is usual to use as primary variables in dimensional formulae

for things with which we are concerned length denoted by /, time

denoted by t, and mass denoted by m.
Since, however, velocity, denoted by v, is proportional to or t

is proportional to
we may use m, I and v as primary variables.

Further, if, as in the Law of Comparison, we assume certain

relations to exist between / and m and / and v, we can express di-
mensional formulae in terms of / alone. For the Law of Compari-
son we assume m to vary as l and v z
to vary as /*. The table
below gives the dimensional formulae of importance for our

Dimensional Formulae.

In I alone when
Law of Com-
parison rela-
In m. /, /. In m, I, v.
tions between
m, I and v

Area or surface ft

Angular velocity and revolutions per i

minute J

Angular acceleration p I

Linear velocity V V7

Linear acceleration

Moment of inertia ml2 mft

Momentum mv

Moment of momentum or angular) mp mvl

momentum j

Force or resistance _

Work, energy and torque



Pressure or stress per unit area.

It will be observed that the relations in the third column agree

with those deduced in various specific cases when considering the
Law of Comparison.

4. Wetted Surface
i.Importance of Surface Resistance. For all but a minute
proportion of actual steam vessels the skin friction resistance, or
the resistance due to friction of the water upon the immersed hull

surface, is greater than the resistance due to all other sources of

resistance combined. For some of the fastest Atlantic liners, for

instance, the skin resistance at top speed, under ordinary smooth-
water conditions, is about 64 per cent of the total resistance. For
only the comparatively few vessels that are pushed to a speed very
high in proportion to their length does the residuary resistance
due to all causes surpass the skin resistance.
Such extremely fast vessels are nearly all for naval purposes.

They are seldom warranted

by commercial conditions.
In view of the great importance of the Skin Resistance it is

advisable to make a careful investigation into the question of the

wetted surface of ships. We need to know how to calculate it

accurately, and how to estimate with close approximation. We


need, too, if the question of wetted surface is to be given its proper

influence in design work, to understand the relations between
wetted surface and size, proportions and shape of ships.
2. Appendage Surface. The wetted surface of hull append-
ages can be calculated as a rule without difficulty. Appendages of
importance have nearly always plane or nearly plane surfaces, and
their areas are readilydetermined by straightforward processes.
Appendage surface, then, can be calculated by simple methods,
the exact procedure varying with circumstances. In dealing with
such appendages as bilge keels and docking keels, which cover or
mask some best to deduct
of the surface of the hull proper, it is

from the gross area of the appendage the area masked by it, the
net area resulting being the addition to the wetted surface of the
hull proper due to the presence of the appendage.
3. Surface of Hull Proper. When we undertake the accurate
calculation of the wetted surface of the hull proper of a ship, we
encounter at once a serious difficulty. It is not possible to develop
or unroll into a single plane the curved surface of a ship's bottom.
We can draw a section at any point and measure its girth, and if
the ribbon of surface included between two sections a foot apart
were equal in area to the girth in feet of the section in the middle
of the ribbon, it would be very simple to determine accurately the
wetted surface of the hull proper by applying Simpson's Rules or
other integrating rules of mensuration to a series of girths at equi-

distant stations, covering the whole length of the ship. Unfortu-

nately, however, on account of its obliquity, the area of this ribbon
of surface general appreciably greater than its mid girth, and
is in

for the best results we must devise a more accurate method. The
simplest plan is to correct the mean girth in question, multiplying
it by a suitable factor, so that the area of the ribbon will be equal
to the corrected mid girth. Then we can apply
the ordinary rules
to the corrected mid girth and obtain accurate results. Let us see
now how to determine the correction factor first for one point of a

section and then for a whole section.

4. Obliquity Factors. In Fig. 31 suppose AB, drawn straight

for convenience, to represent a short portion of a section of a ship's
surface by a normal diagonal plane. CD is parallel to the fore and
aft line. Let AB cut the section FE in E and adjacent parallel
sections each six inches from at L and K.FE Fig. 32 shows dia-
grammatically the three sections and the diagonal plane on the
body plan. The oblique line KL is an element of surface, and we
want to connect its length with ML, the distance between stations.
Now, KL = ML sec KLM. Hence sec KLM is the factor we
need. Now tan KLM
KM = mk
r^TT,, = - - in Fig. 12 T
In practice, then,
if we take a point on a section midway between two other or end
sections, draw a line on the body plan at the point perpendicular to
the section and measure the intercept (mk in Fig. 32) between the
two end sections, we have
Tangent of angle of obliquity =
distance between end sections
and correction factor for obliquity at the point = secant of angle
of obliquity.
Wedo not want to calculate tangents and secants, and we wish
to work directly from the body plan. So we divide the sections
on the body plan at six points into five equal parts. The most
satisfactory method is to lay off small chords with a pair of dividers
and thus determine the points of division by trial and error. Then
we prepare a paper scale so divided that when set perpendicular
to a section at a division point we read at once the correction factor
for obliquity from the intercept between the two sections adjacent

to the one for which we are determining obliquity factors. The

paper scale can be laid off graphically, but can also be readily
calculated. Let us suppose that the actual distance apart of suc-
cessive sections in the sheer or half-breadth plan is i inch. Then
the distance between two sections on either side of a middle section
will be 2 inches. Suppose at a certain point the intercept of the
perpendicular in the body plan between the two stations adjacent
to the one we are considering is 0.25 inch. Then the tangent of
- =
the angle of obliquity is .125. Hence at this point the

angle of obliquity is 7 y'i since tan" 1 .125

= 7 7' \. The
correction factor at i or 1.00778.
the point is Then
sec 77
for our scale \ inch corresponds to a correction factor of 1.00778.
But to lay off our scale we want to determine the varying lengths
corresponding to equal intervals of correction factor.
The necessary calculations are shown in Table II which applies
directly to i-inch section spacing.
Of course, an actual set of lines would nearly always have sec-
tions spaced more than i inch on the plans. For instance, a ship
416 feet long between extreme stations, with 21 stations or 20
spaces, would, if the plans were on the scale of \ inch to the foot,

have the sections on the plans spaced - X - = 5.2 inches. For

20 4
such a ship the data for laying off the proper obliquity scale would
be obtained by multiplying the figures in Column 4 of Table II
by 5.2.

Sample Calculations.
5. Fig. 33 shows an actual body plan
with each section divided into five equal spaces for the purpose
of measuring obliquity and an obliquity scale in place measuring
a correction factor of 1.015 for a point on section No. 15. Table I
shows the calculations in standard form. It is seen that for each
section the average correction factor for obliquity is calculated
from the measurements at six points. The actual measured mean
girths having been corrected, the wetted surface is readily calcu-
lated. The trapezoidal rule is used for the work, being really as
accurate as Simpson's for curves of the type to be handled, and
much shorter.

6. Average Correction Factors. It is seen that the correction

factors for obliquity are always very close to unity. Advantage
may be taken of this fact when dealing with ships of ordinary form
to utilize average correction factors which, when multiplied into
the product of the mean girth by the length, will give the wetted
surface with great accuracy, i.e., within a small fraction of one per

Fig. 34 gives contour curves of correction factors for obliquity

- or ratio between length and beam, and
plotted upon values of >


> or ratio between length and draught.

For vessels of ordinary form it will be found that by determin-
ing the mean girth and applying the correction factor from Fig. 34
the wetted surface is determined with substantially the same
accuracy as if complete calculations had been made. Fig. 34 must
be used with caution for vessels not of ordinary form, if very
accurate results are wanted.
7. Girths of Sections. Having seen how to determine with

accuracy the wetted surface of a ship of which complete plans are

available, I will now take up the determination of the approximate
wetted surface of a vessel whose dimensions and displacement
are known, but for which complete plans are not yet available.
This is a calculation which must often be made. Consider
first the question of the girth of a ship section below water.
This varies with dimensions, proportions, and shape or fullness of
section. The variation with dimensions is a very simple matter.
For similar sections the girth varies as any linear dimension,
such as beam, or draught or Varea. It is convenient to use
Varea as governing quantity and express the girth G of a section of
area in square feet = A by G = g V'A For all similar sections of

varying dimensions the quantity g in the formula preceding is

constant. It is in fact the girth of a section of one square foot area
and similar in all respects to the section whose area is A. Being a
measure as it were of the girth, let it be called the girth parameter.
We want now to ascertain how the girth parameter of a section
varies with proportions and shape. The girth parameters of a

few simple sections are obvious. Thus, if we have a square sec-

tion of one square foot area, the beam is equal to the draught and
the girth is 3 feet, or the girth parameter is 3. If the section is
rectangular of \ foot draught and 2 feet beam, the girth parameter
isagain 3. We
can in fact express by a formula the girth parameter
of a rectangular section of any proportions. Let denote its B
beam, xB its draught. Then xB 2
is its area A and 2 xB the ,
G = -- 2 XB = I + 2 X
Now *u parameter g =
*v r> AT *u -=
girth G. the girth =

Fig. 35 shows a curve of girth parameter for rectangular sections

plotted on x. The minimum value is 2.8284 for x = %, for which,
if the section is one square foot in area, the beam is 1.4142 and the

draught is .7071. For a semicircle of radius r the area = - -


the girth irr. Whence g = -- =\/Tr = 2.5066. This value

2.5066 for a semicircle appears to be the minimum girth parameter
possible. The sectional coefficient for a circle is .7854, and, as
will be seen, this coefficient is close to that for a minimum girth for
any proportion of beam and
8. Actual Girth Parameters. The best to investigate the
variation of girth parameter with proportions and fullness of
section is to draw a number of sections of varying proportions
and fullness and determine and plot their girth parameters. This
has been done for a large number of sections covering a wide range
of fullness and proportions. These sections were all calculated
from the same basic formula, the variations of fullness, etc., being
obtained by variation of coefficients. The details of the work are
somewhat voluminous and need not be given. The results are
fully summarized in Fig. 36, which gives contour curves of girth

parameter plotted upon values of and sectional coefficient.

Fig. 36 is not, of course, applicable to freak or abnormal sections,
but throughout its range is believed to be practically exact for
sections of usual type.

For instance, Fig. 37 shows a series of sections of which No. i

isa parabola and No. 6 is made up of two straight lines and the
quadrant of a circle. The other four sections divide into five equal
parts the intercepts between i and 6 of diagonal lines through O.
Four other figures similar to Fig. 37, except that they had different

proportions, were drawn, and the areas and girth parameters of

the 30 sections thus obtained were carefully determined. Table III
shows these actual girth parameters and girth parameters for the
same proportions and fullness as taken from Fig. 36. The actual

girth parameters were calculated to the nearest figure in the third

place only.
It is seen that Fig. 36 applies to the curves of Fig. 37 and the
other derived figures with great accuracy.
As instancing its application to actual ships' sections attention is
invited to Table IV. This gives for 20 actual midship sections of
vessels whose dimensions and proportions are stated, the actual

girth parameters as measured and the girth parameters from

Fig. 36 for sections of the same proportions and coefficients. The
agreement is very close indeed.
It is evident from Tables III and IV that Fig. 36 represents
with great accuracy the variation of girth parameters of usual
sections of ships as dependent upon ratio of beam to draught and
coefficient of fullness.It follows that, substantially, these are the

only variables. That is to say, if we settle the beam, draught and

area of a section of usual type, we substantially settle the girth,
which varies but little with possible changes of shape. Of course,
this does not apply to sections that are very hollow, having coeffi-
cients well below .5. Fig. 36 does not cover such sections, nor
sections of extreme proportions of draught to beam, such as for-
ward and after deadwoods. For such sections the girth parameters
vary with great rapidity for small changes of beam. Fig. 36,
however, covers nearly all the sections of actual ships of usual
form and is worthy of careful study. We see from it that there
is an actual minimum girth parameter a little greater than 2.5

occurring for
= 2 and coefficient of fullness a little below .8.

Probably we may safely call the coefficient for minimum girth


parameter .7854, the coefficient for a circle. Roughly speaking,


as we vary the minimum girth parameter is always found for

sectional coefficient in the neighborhood of .8 until we get to low

values of below 1.5, where the minimum girth parameters

H ,

correspond to larger coefficients. Similarly, as we vary sectional

coefficient only the minimum girth parameter corresponds very

closely to
= 2 until we reach coefficients greater than when it
H 75

corresponds to smaller values of The most striking feature of

Fig. 36, however, is the comparatively small variation of girth

parameter over a range of values of and sectional coefficient

which covers the bulk of the sections of actual ships. This fact
is of great importance in connection with the determination of a
reliable approximate formula for wetted surface and the considera-
tion of the influence of dimensions, proportions and shape upon
wetted surface.
9. Approximate Formula for Wetted Surface. Suppose we take
n + i sections of a given ship, equally spaced at n + i stations
o, i, 2, 3 ... n. For each section, with subscript denoting the
station, denote the girth by G, the girth parameter by g and the
area by A . Let L denote the length and G the mean girth. Then

= go\/A , GI= gi VAi and so on.
Using the trapezoidal rule we have

Let S denote the wetted surface. Then neglecting obliquity, which
will take care of itself later, when we determine coefficients from
actual ships, we have


If we keepsame sections and space them twice as far apart,

we double length and displacement. We also, neglecting obliquity,
double the wetted surface. If we keep length the same and double
the area of each section, we double displacement. The girth param-
eters of the individual sections are unchanged, so that the result
is to multiply S by va. Now, what convenient
expression in-
volving only length and displacement will give us the same varia-
tion? Evidently, if we write S = C \/DZ, where is displacement D
in tons, L is mean immersed length in feet and C is a coefficient
depending upon proportions, shape, etc., but not upon dimensions,
we have an expression for S which will vary for similar vessels just
as the almost rigorous expression deduced above. For, if we double
length and displacement, we double we keep L constant and
double D, we multiply 5 by W 5; if

As regards primary variation, then, this expression is as accurate

as the rigorous one. It should be carefully noted that L in this
formula is the mean immersed length, or the average water line
length. In many types of vessels the water line lengths are suffi-
ciently close to the mean immersed lengths to be used without
error, but in others, the stem and stern profiles are such that for
accurate work the mean immersed lengths must be determined.
For rough work and first approximations before we are in a position
to determine from plans the mean immersed length, load water
line length is used. Secondary variation in the rigorous expression
given above can come only with variations of the girth parameters,
go, gi, etc. The principal factors affecting the girth parameters
are, as we have seen, variations of ratio of beam and draught and
variations of sectional coefficient. Our formula S = C \/DL so
far takes no direct account of these. They will show themselves
in variations of the coefficient C from ship to ship.
10. Variation of Wetted Surface Coefficient.
Consider, first,
the effect upon wetted surface coefficient of the ratio between
beam and draught. This variation is most conveniently referred

to the value of for the midship section. Fig. 38 shows the varia-
tion of wetted surface with the variation of for the lines of the

United States Practice Vessel Bancroft. Keeping length and dis-

placement constant, a number of body plans were drawn from her
lines with varying from i to 6. The wetted surface for each
ratio was calculated and the resulting curve is shown plotted on
in Fig. 38. It is seen that the minimum wetted surface is found
at = 2.8; but as is changed the variation is slow until we
reach small values of ,
when the wetted surface begins to increase
rather rapidly. Such small values of by the way, are below
H ,

values found in practice. The general features of Fig. 38 could be

inferred from Fig. 36. We see from the latter figure that for a

single section the minimum value of g is found for = 2. Now,

if for the midship section we had = the girth parameter ot
H 2,

this one section would be a minimum, but for every other section
the girth parameter would be above the minimum, since for every

other section would be less than 2. Also for the smaller values
of the girth parameters increase more rapidly than for the larger
values. Henct, for actual ship lines of given length and displace-
ment, but varying the minimum wetted surface must correspond
H ,

to a value of than and the wetted surface would in-
t? H greater 2,

crease, of course, on each side of the minimum. This minimum

is found at = 2.8 in Fig. 38.
It is not so easy to connect the variations of girth parameter of
an actual ship with variations of sectional coefficient. Further-
more, Fig. 36 shows such small variation of girth parameter for
sectional coefficients ranging from .7 to .9 that we may expect to

find in practice the variation due to sectional

masked coefficient

by other arbitrary causes impossible to reduce to rule, such, for

instance, as unusual amount of deadwood or extreme reduction
of deadwood.
However, broadly speaking, the fuller the midship section, the
fuller all the sections are likely to be, and, if the midship section is

very are likely to be fine.

fine, all sections These principles con-
sidered with Fig. 36 would lead us to expect in practice, when using
the formula S = C \/DL, to find rather high values of C associated
with very fine midship sections, and possibly a minimum value of C
for a fairly high midship section coefficient.

In this connection attention is invited to Figs.

39 and 40, which
show variation of wetted surface coefficient with midship section

coefficient, Fig. 39 for fine ended models and Fig. 40 for full ended
models. The four curves in each figure refer to different values
/ L \

of the coefficient Z> -h -las indicated. The higher values of


wetted surface coefficient are found with the higher values of the

/- L- \ 3
coefficient D -f-
I This is to be expected, since the greater
the displacement on a given length the greater the obliquity.

Figs. 39 and 40 refer to a single ratio of beam to draught, namely

2.923, but they show distinct minimum values of wetted surface
coefficient in the neighborhood of midship section coefficients of

.90. As regards absolute values of the coefficients it is to be noted

that at midship section coefficient .84 they are practically coincident.
For higher values of the midship section coefficient the fine ended
models have the smaller wetted surface. For smaller values of
midship section coefficient the fine ended models have the greater
wetted surface. The extreme variations of coefficients in Figs. 39
and 40 are but about 3 per cent above and below the average, a
fact which shows that the coefficient C in the approximate formula
is nearly constant in practice.
ii. Average Wetted Surface Coefficients. Figs. 39 and 40
refer to models of only two types of lines.

A large number of actual wetted surfaces for many types of lines


have been calculated at the model basin from which Fig. 41, show-

ing contour curves of the wetted surface coefficient C plotted on

and midship section coefficient, has been deduced.
The wetted surface coefficients of Fig. 41 were obtained from

average results of vessels of ordinary form. For such vessels, if

the mean immersed length is accurately known, they are correct
within a small percentage. They apply to the hull proper only,
exclusive of appendages, and should be used with caution for vessels
of abnormal form, such as very shallow draught vessels, vessels
with very broad, flat sterns, vessels with deadwood cut away to an
unusual extent, etc.
In practice Fig. 41 can be utilized to ascertain with a good deal
of accuracy the wetted surface of a vessel of abnormal type, provided
we have the correct value of C for one vessel of the type which does
not differ too much in proportions and coefficients from the vessel
whose wetted surface is needed.
For, suppose that Fig. 41 is 4 per cent in error for the abnormal
vessel whose wetted surface coefficient is known. It will continue
to be very approximately 4 per cent in error for the type of lines
under consideration as proportions and coefficients are changed,
and its results corrected by 4 per cent may be relied upon for the
abnormal type. In other words, Fig. 41 may be utilized in two
a. To ascertain the approximate wetted surface of any vessel
of ordinary type whose dimensions, displacement and midship
section area only are known.
b. To ascertain the approximate wetted surface of a vessel of

extraordinary type of known dimensions, etc., provided we know the

actual wetted surface of another vessel of the same extraordinary

From a consideration of what has gone before, and especially
of Figs. 36 to 41, we appear to be warranted in drawing a few broad
conclusions as to the wetted surface of vessels of usual types.
1. For a given displacement the wetted surface varies
with length, being nearly as the square root of the length.
2. For a given displacement and length the wetted surface varies

but little beam and draught possible in practice.

within limits of
As regards wetted surface the most favorable ratio of beam to

draught is a little below 3.

3. For given displacement and dimensions the wetted surface is

affected very little by minor variations of shape, etc. Extremely

full sections are somewhat, and extremely fine sections are quite

prejudicial to small surface.

4. After length, the most powerful controllable factor affecting
wetted surface is probably that of deadwood. By cutting away
deadwood boldly, we can often save more wetted surface on a ship
of given displacement and length than by any practicable variation
in ratio of beam to draught, or in the fullness of sections.

5. Focal Diagrams

1. Field for Focal Diagrams. In attempting to analyze experi-

mental data itfrequently happens that we know the general law
which we think should govern, and we wish to examine whether
the law does apply and, if it does, to determine suitable coefficients
from the experiments for use in the formula expressing the law.
Experimental data being at best an approximation, it is desirable
to use a method which will not only give us an adequate approxi-
mation to the coefficients or constants desired, but give us some
idea as to how closely our results are going to represent the ob-
served data.
Mathematically, the problem is in general one of Least Squares.
In practice, for many problems there is one coefficient or constant
to be determined, the actual determination, of course, being made

by taking average results. In a great many cases not so simple

there are two coefficients or constants involved. For such cases,
instead of applying the complicated and laborious methods of
Least Squares, very satisfactory results can always be obtained
from data not too much in error by the use of what I may call a
Focal Diagram.
2. Illustration of Focal Diagrams. This method may be
readily comprehended from a concrete illustration. Fig. 42 shows
a parabola whose equation is y = 3x ,
the general equation

being of the form y

= ax bx~. At the point P, say, where
x = 4, y
= 8. Substituting these values of x and y in the general
equation, we have 8=40 i6&. This is a linear relation between
a and b, and laying off axes of a and b as in Fig. 43, we can draw
a line representing this relation. If we take the simultaneous
values of a and b for any point on this line and substitute them in
the general formula y = ax bx2 the resulting parabola in x and

y would pass through y = 8, x = 4.

Fig. 43 shows ten lines in a and b corresponding to ten points
on the parabola y = 3 x These points are as follows:
x,= i 23 45 67 89 10
= 2.75 5 6.75 8 8.75 9 8.75 8 6.75 5

These ten lines all pass through the point a

.25, forming
= 3, b
a focus at this point. Evidently, if we know the x and y values of
the ten points and the fact that they are on a curve whose formula
is of the form y = ax bx2 we could determine a and b by drawing

the ten lines as in Fig. 43 and taking the focal values a = 3, b = .25.
If we knew the exact ordinates of but two spots, we could draw the

two corresponding lines in Fig. 43 and determine the values of

a and b.
In practice,if we determined the spots on the curve by experi-

ment or observation, we would have more spots than theoretically

needed to determine the focus; but the line for each spot instead
of passing through the focus would pass somewhat near it, its
distance from the focus depending upon the nearness of our
observations to exact truth.
In Fig. 42, circles on the curve indicate ten exact spots, and
adjacent crosses indicate spots of varying errors in location. The
errors,both vertical and horizontal, vary by .05 from + .25 to
.25, and the actual errors at any spot were assigned by lot.

We have, then, for the approximate spots

x = i 1.75 2.85 4.10 4.80 5.90 7.20 8.05 9.25 9.95
y = 2.$5 4.85 6.65 8.15 8.75 9.20 8.85 7.75 7.00 5.05
A focal diagram similar to Fig. 43 can be drawn with a line
for each approximate spot, and this is done in Fig. 44. It is

evidently possible in Fig. 44 to spot the focus with an accuracy

ample for most practical purposes.

3. Considerations Affecting Focal Diagrams. If the assumed

law or general equation is materially in error, a good focus will not

be formed, no matter how close the observations may be. Even
with an exact law it may be difficult to locate the focus if the
observations are poor, but when we do get a good focus we know
at once that the corresponding values of the coefficients in our for-
mula will cause the formula to represent the experimental results

with great accuracy, indicating that the assumed formula is close

to the truth and that the observations are good.
In Fig. 44 the lines are straight. This need not necessarily be
the case. The relation between a and b may not be linear, but can

always be represented by a curve. Linear focal diagrams are,

however, much the simplest and best and should always be sought
for. Frequently, when the relation between the coefficients is not
linear, it may be made so by adopting new coefficients of definite
relation to the original ones.
In a linear focal diagram we usually determine two points on
each line. The exact methods best to use vary somewhat with the
nature of the case. It is always desirable to determine the two

points, one on either side of the focus. Below are given the
detailed calculations for the case we have been considering from
the results of which Fig. 44 was plotted.

= ax


by?, a =-
bx, b = o, a =
b = .5, a = z + .5 x.

disturbance of the water and before considering in detail the ele-

ments of resistance it will be well to form some idea of the nature of
the disturbances to which resistance is due.
i. Comparison between Ideal Stream Motion and Actual Motion.
Suppose we could apply on the surface of the water a rigid
frictionless sheet as of ice surrounding to a great distance a moving
ship and advancing with it. If the ship had a smooth and friction-

less bottom and the water were a perfect liquid there would be
perfect stream line motion, and we know from stream line considera-
tions the salient characteristics of what may be called the stream
line disturbance in the vicinity of the ship's hull. In the vicinity
of and forward of the bow the water would be given a forward and
outward motion, with pressure in excess of that of the undisturbed
water. Passing aft, the water would continue to flow outward,
but at a short distance abaft the bow would lose its forward motion
and begin to move aft as well as outward. Its pressure, a maximum
near the bow, would steadily fall off, soon becoming less than that
of undisturbed water.
Abreast the midship section, the sternward velocity would reach
a maximum and the pressure a minimum. Passing sternward, as
the water closed in it would lose its sternward velocity, and pressure
would increase again until in the vicinity of the stern we would have
excess pressure and the water would have motion forward as at the
bow. Since there would be a deficiency of pressure over the greater
portion of the hull, we must, in order to realize the ideal motion,
assume that the rigid sheet surrounding the ship is strong enough
to hold it firmly at the level at which it naturally floats when at
rest. We must also assume that the pressure of the undisturbed
water is such that the defect of pressure caused by the motion of

the ship will not cause the water to fall away from the rigid sheet.
Now the motion of the actual ship through actual water differs
from the ideal conditions assumed above.
1. The water is not frictionless, but is affected by the frictional

drag of the surface of the ship.

2. The ship is not constrained to remain at a fixed but
level, may
rise and fall bodily and change trim in response to the reactions of
the water.

3. The water not constrained to remain at one level,

surface is

but is free to rise and

response to the action of the ship.
fall in

2. Changes of Level of Vessel and Water.

the differences between the actual circumstances of the motion and
the ideal conditions assumed above, there is no doubt that the
stream line action around an actual ship presents in a qualitative

way nearly all the features of the ideal case considered. But in
the actual ship the excess pressures at bow and stern result in
surface disturbances, causing waves which spread away and absorb

energy, and the defect of pressure amidships results in a lowering of

the water level and a lowering of the ship bodily, accompanied by
a change of trim.
Figs. 45 to 49 show for two speeds of one model and three speeds
of another changes of level and trim of model and of level of water

against the side. The dimensions and displacements of the models

are given in the legend just above Fig. 45.
These figures are typical. They show elevations of the water
at bow and stern, and show further two phenomena already de-
scribed as to be expected from stream line action but not conspicu-
ous or easy to determine for an actual ship. It is seen that there
is a bodily settlement of the vessel and that in the vicinity of the
mid length there is a bodily lowering of the water surface adjacent

to the ship independent of the disturbance due to the wave created

at the bow.
3. Lines of Flow over Surface of Vessel. There have been a
number of experiments made at the United StatesModel Basin
upon the direction of relative flow of the water in the vicinity of
models. The model surface being coated with sesquichloride of
iron mixed with glue, pyrogallic acid is ejected at a point of the
bottom through a small hole, which as it passes aft mingled with the
water causes a gradually widening smear of ink upon the prepared
model surface. The center line of this smear can be located with
reasonable accuracy for some distance, and when it becomes uncer-
tain a fresh hole is bored and the line traced on. When experiment-
ing with flow not in the immediate vicinity of the model surface,
meshes of fine string or wire coated with sesquichloride of iron are
used and pyrogallic acid ejected at known points.

The relative flow indicated in the immediate vicinity of the model

is found to extend as regards type quite a distance from the skin,
so as regards motion near the hull we need consider only the dis-
turbance close to the bottom, or the lines of flow as they may be

Figs. 50 to 59 show lines of flow past the bottom

for five pairs of

twenty-foot models of five widely varying types of midship section.

The proportions, displacements and speeds of the models are given.
The large and small models of each type of midship section are
similar except as regards ratios of beam and draught to length.
These figures are typical and confirmed by investigations of the
lines of flow over a number of other models. Perhaps their most
notable feature is the remarkably strong tendency of the water to

dive under the fore body as it were. In fact, it seems as if the water
near the surface forward dives down and crowds away from the
hull the water through which the fore part has passed, while aft
the water rising up crowds away from the hull the water which
was in contact with near the surface amidships.

4. Kelvin's Patterns and Actual Ship Wave Patterns. -

It remains to consider the most striking of the disturbances caused

by a moving ship. This is the surface or wave disturbance.

The wave disturbance caused by a ship differs obviously from
trochoidal waves, which we have considered.
These latter were considered as an infinite series of parallel

crests, each crest line extending to infinity.

We owe to the genius of Lord Kelvin the solution of an ideal

problem which applies reasonably well to ship waves. His work

in this connection, which may be found in the Transactions of the

Royal Society of Edinburgh (Vols. XXV (1904-5) and XXVI

(1906)), bristles with difficult mathematics, but his results are

comparatively simple.
Suppose we have advancing in a straight line over the surface of
a perfect liquid a point of disturbance. What will be the resulting
waves? Lord Kelvin's conclusion is that there will be a number of
crests, each crest line being represented by

_ 20 ^.

where the origin is supposed to travel in the direction of the axis

of x with and at the point initiating the disturbance.
The equation above is somewhat simpler in polar coordinates.
Transformed it becomes

4 - aV2 (i + 18 sin 2 6-27 sin4 6) + 16 a4 sin 2 6 = 0.

shows a single crest line from the above equation. It

Fig. 60
always from o, where it is tangent to the axis of x. It spreads
outward and backward to cusps CC, which are on a line making
with the axis of x the angle of 19 28' whose tangent is Vf or sine
is ^. The tangent at the cusp is inclined 54 44' to the axis of x,
and the branch CAC of the crest line is perpendicular to the axis
of x where it crosses it. The relative heights of various points on
the crest as given by Lord Kelvin are indicated in Fig. 60. The
fact that the heights at and CC are infinite shows simply that
the formula cannot represent the physical conditions with exactness.
Itmay, however, be an amply close approximation, for by the theory
these infinite crest heights extend for but infinitely short distances.
The physical interpretation of the formula is that at OC and C
the heights are greatest and the crests the sharpest, so that at these

points, if anywhere, breaking water will be found. This conclusion

is fully borne out in practice.
The whole wave disturbance due to the initiating point is made

up by the super position of a series of crests such as are outlined in

Fig. 60, with corresponding intervening hollows. Fig. 61 shows a

such crest lines. The diverging crest lines cross the trans-
series of
verse crest lines, resulting in an involved surface disturbance.
The distance between successive transverse crests along the axis
of x the same as the length of an ordinary trochoidal wave travel-

ing in deep water at the speed of the point of initial disturbance.

The heights of successive crests are inversely as the square roots
of distances from the origin.
That Lord Kelvin's solution agrees reasonably well with practical
results is readily shown by careful scrutiny of the wave disturbances
caused by ships and models, which makes it clear that the bow wave
system and the stern wave system closely resemble Kelvin wave

The differences are only such as might be expected from the fact
that a Kelvin group is an ideal system initiated by forces at a single
moving point, while an actual wave group is due to forces spread
over the ship's hull.
heights of the later diverging waves close to the ship appear
to be much less in practice than by the Kelvin formula, these crests

frequently appearing as mere wrinkles of the surface, and the ship

wave patterns vary with proportions of the vessel. Thus narrow
deep ships have wave patterns whose transverse features are much
more strongly accentuated than those of broad shallow ships.
The wave patterns of ships appear to change somewhat with
change of speedand the transverse features appear to be less promi-
nent and important at high speed. According to observations
made by Commander Hovgaard, formerly of the Danish Navy, and

given by him in a paper before the Institution of Naval Architects

at its spring meeting in 1909, the cusp line is usually at an angle
less than 19 28', most observations of full-sized ships showing it
between 16 and 19, though in one case, that of a Danish torpedo
boat, Commander Hovgaard observed a cusp line angle as low
as 11.
Observations made on models by Commander Hovgaard in the
United States Model Basin showed even smaller values of cusp line

angles, particularly at relatively high speeds.

But at such speeds the breadth of the basin is not sufficient to
allow the cusp line determined with accuracy.
For purposes of analysis the most important feature of the
Kelvin wave group the close agreement between its curved trans-

verse crests and a series of transverse trochoidal crests extending

from the cusp line on one side to the cusp line on the other.
5. Havelock's Wave Formulae. Lord Kelvin's wave formulae
given above are for deep water. Dr. T. H. Havelock has developed
formulae for the wave patterns produced by a traveling disturbance
in water of any depth. These will be found in a paper on waves,
the Proceedings of the Royal Society, Vol. 81, 1908.
etc., in
In a paper on Wave-making Resistance of Ships, Vol. 82, 1909,
Dr. Havelock has applied his formulae to produce practical results.
For waves generated by a traveling disturbance in deep water

Havelock's results agree with Kelvin's except that Havelock's

formulae do not require infinite wave heights.
But in shallow water Havelock finds that there
is a critical
which is, in feet per second, VgA, where h
the depth in feet. This

is, by the way, the speed of the solitary wave or wave of translation

by the trochoidal formulas.

As the speed increases up to the critical speed the cusp line angle,
which was 19 deep water, becomes greater and greater until
28' in
at the critical speed it is 90. At this speed the wave disturbance
reduces to a single transverse wave.
Above the critical speed transverse waves cannot exist. Diverg-

ing waves continue however, but instead of being concave the first
one is straight at an angle which decreases from 90 with the axis
as speed increases beyond the critical speed.
The succeeding diverging waves are convex instead of concave.
We observed phenomena accompanying the
shall see later that
motion of models in shallow water are in accordance with Have-
lock's theoretical conclusions.


7. Kinds of Resistance

THERE are several kinds of resistance and usually all are present
in the case of every ship. They will be enumerated here and then
taken up separately in detail.

1. Skin Resistance. In the first place, water is not frictionless.

Its motion past the surface of the ship involves a certain amount
of frictional drag, the resistance of the surface involving an equal
and opposite pull upon the water.
This kind of resistance is conveniently denoted by the term
Skin Resistance. It is nearly always the most important factor
of the total resistance.
2. Eddy Resistance. While Skin Resistance is accompanied
by eddies or whirls in the water near the ship's surface, the expres-
sion Eddy Resistance is used for a different kind of resistance.
The motion through the water of a blunt or square stern post or
of a short and thick strut arm, etc., is accompanied by much resist-

ance and the tailing aft of a mass of eddying confused water. Such
resistance is designated Eddy Resistance. With proper design it

is in most cases but a minor factor of the total resistance.

3. Wave Resistance. A far more important factor, which

though usually second to the Skin Resistance is in some cases the
largest single factor in the total resistance, is the resistance due to
the waves created by the motion of the ship. It is called for brevity
the Wave Resistance.
We have seen that the motion of a ship through the water is
accompanied by the production of surface waves. These absorb
energy in their production and propagation, and this energy is
communicated to them from the ship, being derived from the Wave

4. Air Resistance. we have the Air Resistance, which

is, as its name
implies, the resistance which the air offers to the
motion of the ship through it. The Air Resistance is seldom large.
It is, however, by no means always negligible.
5. Comparative Importance of Skin and Wave Resistance.
Considering the two main factors of resistance, namely, Skin
Resistance and Wave Resistance, experience shows that for large
vessels of very low speed the Skin Resistance may approach 90 per
cent of the total. For ordinary vessels of moderate speed, it is
usually between 70 and 80 per cent of the total. As speed increases,
the Wave Resistance becomes a more and more important factor,
until, in some
cases of vessels pushed to speeds very high for their

lengths, the Skin Resistance may be only some 40 per cent of the
total, the Wave Resistance being in the neighborhood of 60 per
cent. For such vessels as high-speed steam launches the Wave
Resistance may be even more than 60 per cent of the total, but for
vessels of any size it is seldom advisable to adopt a design where
the Wave Resistance as great as 50 per cent of the total.

Features which tend to decrease Wave Resistance tend to in-

crease Skin Resistance, and here, as in so many other matters, the
naval architect must adopt a compromise dictated by the special
considerations affecting the particular case.

8. Skin Resistance

i. William Froude's Experiments. The determination of the

Skin Resistance of ships is based entirely upon the experimental
determination of the frictional resistance of thin comparatively
small planes moving endwise through the water. The classical

experiments in this connection were made by Mr. William Froude

many years ago and are recorded in the Proceedings for 1874 of the
British Association for the Advancement of Science.
Mr. Froude used boards i\ X 19 inches, of various lengths up to
50 feet and coated with various substances, which were towed at
various speeds not exceeding eight knots in a tank of fresh water
300 feet long, their resistance being carefully measured. Mr.
Proude summarized his experimental results in the following table:







This seems natural, and most experiments on the loss of head of

water flowing through pipes show that the resistance to flow varies
as the square of the speed. The conditions are, however, very
different. In the case of the pipe we consider the average velocity
of flow over the cross section of the pipe, which is necessarily the
same from end to end, and its ratio to the rubbing velocity of the
water close to the walls of the pipe is practically constant. In
the case of the plane, the rubbing velocity steadily falls off along
the plane.
While the frictional index 1.83 for long smooth surfaces does not
differ greatly from 2, the corresponding curve is far below the par-
abola corresponding to the index 2. Thus the ratio F 83
1 '


which is unity for V= i, is .761 for V= 5, is .676 for V = 10 and

.609 for V= 20. This ratio falls off more and more slowly as
speed is increased. Thus, in passing from V= i to V= 20 it

falls off from to .609, while to reduce it to .500 the speed must
increase to V= 59.
3. Frictional Resistance of Ships Deduced from Plane Re-
sults. In order to apply the results for friction of planes to the
frictional resistance of ships, it is necessary first to extend the

experimental results for short planes to long surfaces, the lengths

of actual ships. This has been done by Froude and Tideman, by
extending the curves of index, coefficient, etc., for the short planes

experimented with. While this extension is speculative to some

extent, it does not appear that it is likely to be seriously in error.
Then it is assumed that the frictional resistance of the wetted
surface of a ship is the same as the frictional resistance of a plane
of the same length and total surface moving endwise through the
water with the speed of the ship. This assumption is necessarily
an approximation. The water level changes around a ship under
way, changing the area of wetted surface; and, owing to stream
over the surface is at some places
line action, the velocity of flow

less, at others greater, thanwould be over the plane surface.


The assumption made, however, is practically necessary, and is a

reasonably close approximation to actual facts.
Finally, it is necessary to assume that the frictional quality of
the ship's surface is the same as that of our experimental planes.

From experiments made in the Italian model basin and else-

where it may be concluded that the frictional resistance of a
smooth hard surface is not materially affected by the variety of
paint with which it is covered. But Froude's experiments show
that friction powerfully affected by roughness of surface. For

a 5o-foot plane covered with calico or medium sand and towed at

600 feet per minute, or about 6 knots, Froude found a frictional
resistance nearly double that of a varnished plane of the same
size. The calico surface had an index but little greater than the
varnished surface, so its friction would remain in nearly constant
ratio to that of the varnished surface. The medium sand, how-
ever, had a greater index. This results in a much greater rela-
tive increase at high speeds. Thus, using Froude's coefficients,
the ratio between medium sand and varnish, which is 1.43 at one
knot, becomes 2.12 at 10 knots, 2.38 at 20 knots, and 2.56 at 30
The relatively enormous increase of. frictional resistance with
fouling is well known, but we have very little quantitative infor-
mation as to the difference as regards frictional quality even be-
tween the smoothest possible steel ship and one whose bottom,
while acceptably fair, is not ideally smooth.
It would be very desirable to narrow the gaps which we must
now bridge by assumptions in connection with frictional resist-
ance from the results of experiments on large and long planes of
various surfaces made in open water at high speeds. Such ex-
periments would, however, be very difficult. It would be very
hard to tow such planes straight.
Pending such experiments, we must rely upon coefficients
deduced from the small scale experiments.
4. R. E. Froude's Frictional Constants.Mr. R. E. Froude, in
a paper in 1888, before the Institution of Naval Architects, has
supplemented the British Association paper of his father, Mr.
William Froude, by data of coefficients and constants used by him,
from which Table V of Froude's Frictional Constants has been
It will be noted that as regards paraffin surfaces the table
differs slightly from Mr. William Froude's results, obtained in 1872.

Mr. R. E. Froude states that as regards the paraffin in use in

1888 appeared identical in frictional quality with a smooth

painted or varnished surface.

5. Tideman's Frictional Constants. Closely following the elder
Froude's classical experiments of 1872, Herr B. Tideman, Chief
Constructor of the Dutch Navy, made a number of similar
experiments, from which he deduced a complete set of frictional
constants. These are given in Table VI. The most important
are those for Iron Bottom Clean and Well Painted." These
are comparable with Froude'c constants, and it will be noted that
they are slightly greater.
For varnished planes 20 feet long, Froude's constants agree
very closely with results of careful experiments at the United
States Model Basin; but for full-sized ships it is considered pref-
erable to use Tideman's coefficients, simply because they are slightly

larger, and hence make some allowance for the imperfections of

workmanship found in practice. At the United States Model
Basin, the practice, when dealing with vessels more than 100
it is

feet long, to use the Tideman values of /, but the index 1.83
instead of 1.829, as given by Tideman. This increases Tideman's
results by negligible amounts.
6. Law of Comparison not Applicable to Frictional Resistance.

Having concluded, then, that we should represent the fric-

tional resistance of a ship by Rf = fSV 1 83'

where Rf is frictional
resistance in pounds,/ is a coefficient varying slightly with length,
S wetted surface in square feet and V is speed in knots, let us

see whether we can apply the Law of Comparison to resistance

following the formula.

Let Rif,f\, Si, Vi refer to one ship, R f S V
2 /, z, 2, 2 to a similar ship.
Then R\/= /iSiFV' 83RZ/= fzSzVz .
1 ' 83
. Let the ratio of linear
dimensions of the two ships be X and let F2 and V\ be in the
ratio Vx, as required by the Law of Comparison.


Now = X2

and we have made ~ = V\ = X*.


Then at corresponding speeds

1.83 f

2 = J* X2-915.

But to satisfy the Law of Comparison we should have at corre-

spending speeds -^ = X3 . We see, then, that frictional resistance
does not follow the Law of Comparison, and hence we cannot
deduce the frictional resistance of a full-sized ship from that of
a model. Thus suppose we had a vessel 500 feet long of 12,500
tons displacement and 39,000 square feet wetted surface. A
similar 2o-foot model would have 62.4 square feet of wetted
surface. If the speed of the ship were 20 knots, the correspond-
/ 20
ing model speed would be 4 knots
= 20 yV
Using Froude's coefficient and 1.83 index, the frictional resist-
ance of the 2o-foot model would be .01055 X 62.4 X 4 1-83 =
8.3218 pounds. If the Law of Comparison held, this would make
Rf for the full-sized ship at 20 knots 8.3218 (25) 3 = 130,028 pounds.
But using Froude's coefficient of friction we have for the full-
sized ship Rf = 39,000 X .00880 X 2O1
' 83 = 82,495 pounds, and using
Tideman's coefficient Rf = 84,745 pounds.
It is seen, then, that the Skin Friction, as we calculate it, falls

far short in practice of what it would be if the Law of Compari-

son were applicable to it.

7. Air Disengaged around Moving Ships. There is one phe-

nomenon generally accompanying the motion of a full-sized ship
which seldom manifests itself in model experiments. As a fast
ship moves through the water, it is seen that the water in the
immediate vicinity of the skin plating, particularly aft of the
center of length, has a great The air is either
many air bubbles.

disengaged from water in which by the reduction of

it is entrained
pressure in frictional eddies, or it is carried down and along the
ship as a result of breaking water toward the bow. However

produced, its presence must reduce the density of a layer of water

covering a large portion, if not all, of the surface of the bottom,
and it would seem, at first, that there should be a corresponding
reduction of friction. It is in fact a favorite dream of inventors
to deliver air around the outside of a ship so that the immersed
surface will be surrounded by a film of air instead of water. Could
this result be accomplished, it would undoubtedly result in a great
reduction of skin friction. But air released under water persists
in forming into globules, not films. Experiments have been made
at the United States Model Basin by pumping air around a model

through a number of holes near the bow and out through narrow
vertical slots in the forward portion of a 2o-foot friction plane.
The these experiments were that for the model the
results of
resistance was always materially increased when the air was
pumped out. In this case the air came out through holes and
promptly formed globules. In the case of the friction plane the
air came out in a thin film which spread aft. At speeds of 12 to
1 6 knots, when the films of air on each side visibly extended over

perhaps a third of the plane, the resistance was almost exactly the
same as when no air was pumped. At speeds below 12 knots the
resistance was greater when the air was pumped.
It is possible that for vessels of the skimming-dish or other
abnormal type the efforts of inventors to reduce resistance by

means of air cushions may be successful, but there is little doubt

that no matter how much air may be forced into the water around a
ship of ordinary type, practically none of it remains in contact with
the ship's surface. That is covered always by a film of solid water.
The air forms globular masses or bubbles and never touches the
surface of the hull. While in an actual ship the air bubbles
naturally appearing must somewhat reduce the density of some of
the liquid around the bottom, it appears likely that, to reduce
skin friction materially, this reduction of density would have to
extend to a much greater distance from the hull than is usually the
case and that in practice the evolution of air found probably in-
creases the resistance by an uncertain amount. This uncertainty
could be removed by friction al experiments upon planes of such size
and nature of surface as to be closely comparable to actual ships.

8. upon Skin Resistance.

Effect of Foulness In design work
we usually deal primarily with clean bottoms. When vessels
become foul by the accumulation of marine growths such as grass
and shellfish the Skin Resistance is much increased. Fig. 62
change of surface upon Skin Resistance.
illustrates the effect of
Froude's experimental results for five surfaces are extended by
his formula to high speeds. The two smooth hard surfaces
varnish and are nearly the same.
tinfoil But a surface covered
with calico shows about double as much resistance, and surfaces
covered with fine or medium sand show more than double the
resistance of the varnished surface at speeds above 20 knots.
When we reflect that in the most extreme cases of fouling a ves-
sel's bottom may have a complete incrustation of shellfish it is

easy to realize that fouling may result in Skin Resistance four or

five times that of the clean ship.

Of course in practice such fouling is permitted only under ex-

ceptional circumstances, vessels in service being docked at inter-
vals. But even in cool waters where fouling usually goes on
rather slowly a vessel three or four months out of dock is liable to
have an increase of 20 per cent or more in Skin Resistance, and in

tropical waters the increase of resistance is greater.

Foulness usually gauged by the loss of speed, which tends to

mask the great increase of Skin Resistance. Thus a loss of two

knots of speed for the same power means in the case of a vessel
originally of moderate speed an increase of about 100 per cent in
Skin Resistance.
When work it is necessary to allow for the effect of
in design

fouling usually done indirectly by providing a margin of

it is

speed with a clean bottom equal to the loss to be expected from

fouling. This loss must be estimated from previous experience
with vessels in the service under consideration.

9. Eddy Resistance
As already stated, Eddy Resistance is a minor factor in the
case ofmost ships and cannot be determined separately by ex-
periment. It is possible, however, to get a reasonably good idea

of the laws of Eddy Resistance by experiments with planes, sec-

tions of strut arms,and similar appendages.
1. Flow Past a Thin Plane Producing Eddy Resistance. Fig.
63 shows a section through a plane AB and a stream of water
flowing past it, and indicates, diagrammatically, what happens.
The plane is inclined at an angle a to the direction of undisturbed
flow; K
is the dividing point of the stream. On one side of K
the water flows around the corner at A. On the other side it

flows by B. The position of K depends upon the angle a. In

front of the plane there is practically perfect stream motion, as
indicated. The velocity of the water is checked, with corre-
sponding increase of pressure, but there is no discontinuity. In
the rear of the plane, however, the conditions are different. The
water breaks away at A and B, and there is found behind the
plane a mass of confused eddying water, whose pressure must be
reduced below the normal pressure due to depth below the sur-
face, but in a more or less erratic manner.
2. Rayleigh's Formulae for Eddy Resistance. The total Eddy
Resistance of the plane would then be due to a front pressure and
a rear suction. These are evidently but little dependent upon
each other. Thefront pressure has been investigated theoreti-

cally by assuming a smooth solid inserted behind the plane, so that

the water has perfect stream motion throughout. The resulting
formulae as deduced by Lord Rayleigh are as follows :

2 TT sin a. w
n .

4 + sin a 2 g TT

AK _ 2+4 cos a 2 3
cos a + (if a) sin a
AB 4 + TT sin a
In these formulae Pn '
is normal pressure or total pressure per-
pendicular to the front face of the plane, a is the angle the plane
makes with the direction of motion, w
the weight per cubic

foot of the water, g is the acceleration due to gravity, A is area of

plane in square feet and v is its velocity in feet per second.

It may be noted that at K, where the water is brought com-
pletely to rest, the excess pressure is v . If this pressure were

over the whole plane, the total normal front pressure would be
W A 2
Av .

The fraction - is, then, the ratio between the front pres-
a 4 + TT sin

sure and the pressure due to velocity multiplied by the area of the

plane. This fraction is, as might be expected, a maximum for

a = go . Its value, then, is or .88. This is materially less

4 + 7T

than unity, and as a decreases the fraction soon begins to fall off

rapidly. Fig. 64 shows curves of the ratio and the
4 + TT sin a
ratio - plotted on a.

The by Rayleigh's formula follows the Law of

front pressure

Comparison. For suppose we have two similar planes at the

same angle. If P\ denote the front pressure on No. i and P 2 the
front pressure on No. 2,

2 TT sin a w 9 n 2 TT sin a w
4 4- TT sin a 2 g 4 + ?r sin a 2 g

Whence - '-*
Now if X denote ratio of linear dimensions,

AI= \2 A% and for corresponding speeds Vi z = \v z 2 Then at corre- .

spending speeds -=^
= X3 or Froude's Law is satisfied. ,

For salt water -- = i practically. Furthermore it is desirable
to reduce ail speeds to knots, denoted by V. When this is done
Rayleigh's formula for front face pressure may be written

5-705 sin
/y_ + sm a

Joessel's Experiments and Formulae for Eddy Resistance.

When we come to consider the total normal resistance of an in-
clined plane moving through water we are compelled to rely upon
semi-empirical formulae derived by experiments.
It is impossible to reduce the resistance due to confused eddy-

ing behind the plane to mathematical law. The ground has never
been adequately covered experimentally, and it is as a matter of
fact a question whose accurate experimental investigation pre-
sents many difficulties.

M. Joessel made experiments with small planes 12 inches by

16 inches in the river Loire at Indret, near Nantes, about 1873.
The maximum current velocity was only about 2? knots. Joes-
sel's results may be expressed as follows:
If denote the breadth of a plane in the direction of motion

making the angle a with the direction of flow and x the distance
of the center of pressure from the leading edge,

x = (.195 + .305 sin a) I.

If Pn denote total normal force due to pressure in front and

suction in rear, we have for area A in square feet and velocity V
in knots

7.584 sin a . n ,
* n _
, AY ,

.639 + sin


4. John's Analysis of Beaufoy's Eddy Resistance Experiments.

- Mr. A. W. "
John in an interesting paper on Normal Pressures
on Thin Moving Plates," before the Institution of Naval Archi-
tects in 1904, has analyzed Colonel Beaufoy's experiments of

1795 with square plates of about three square feet area (double
plates abreast one another about 8 feet apart and 3 feet below the
surface) and shown that the results present the following peculiar
features. Up to about 30 degrees inclination the normal pressure
increases linearly, and from 30 degrees to 90 degrees it remains
almost constant. The same result has been found by various
recent experiments with planes in air. It appears to be charac-
teristic of squares, circles and rectangles approaching the square,
and is not so pronounced in the case of long narrow rectangles
moving perpendicular to the long side.

Beaufoy's results as plotted by John may be approximately

expressed by a semi-empirical formula of the same form as Ray-
leigh's formula,
~ DA
V '
B + sin


This may be made to coincide at two points with the experimen-

tal results. We have
For coincidence at a = 90 and a = 10 A = 5.20,B = .557
For coincidence at a = 90 and a = 15 A = 4.63, B
= .389
For coincidence at a = 90 and a
= 20 A = 4.08, B = .223
It is reasonable to take the values for a = 15. We then have
formula derived from Beaufoy,

4-63 sin a . ,
Pn _ " '

.389 +

sm a
5. Stanton's Eddy Resistance Experiment. Dr. T. E. Stan-
ton has recently made experiments with very small plates of
2 square inches area in an artificial current of water of 4 knots

velocity. His results are published and discussed in a paper of

April 2, 1909, before the Institution of Naval Architects. He
found the same phenomenon developed by John's analysis of
Beaufoy's experiments, namely that the normal pressure on a
square plate rises almost linearly to an angle of 35 or so and
then does not change much from 35 to 90. For a plate whose
length in the direction of motion was twice its width there was
a pronounced at about 45, the normal pressure at this
inclination being 13 or 14% greater than at 90. For a plate of
length in the direction of motion but one-half its width the hump
feature was not so pronounced and was strongest at an inclina-
tion below 30.
6. Formulae for Eddy Resistance of Normal Plates Compared.
When a = 90, or the plane moves normally to itself, we have
Pressure on front face = Pn = 2.51 A F2

By Rayleigh's formula:
By JoessePs formula: Total normal force = P n = 4.63 A V2
By formula from Beaufoy's results, Pn = 3.33 AV2
From Stanton's results, Pn = 3.42 AV2
It is probable that Rayleigh's formula expresses quite closely
the resistance of a square stem for instance. If we adopt Joessel's
formula, which gives the largest resistance, and deduct the front
face pressure, we would have for rear suction Pr = 2.12 A V

This formula will probably give an outside value for resistance

such as that of a square stern post.


7. Formulae for Eddy Resistance of Inclined Plates Compared.

For small values of a it is convenient to use a formula of the
form Pn = C sin a A V2 If we choose C to correspond to Pn

from the complete formula for an angle of 15 degrees we can

simplify Rayleigh's formula, etc., for use up to angles of 30 or so.
Stanton's results are already expressed in this simple form, and
William Froude has a formula of this type expressing normal
force for small angles of inclination.

Pn = a AV2

Rayleigh's formula becomes 3.73 sin

JoessePs formula becomes Pn = 8.45 sin a A V2
Formula from Beaufoy becomes Pn = 7.15 sin a AV2
Froude's formula becomes Pn = 4.85 sin a AV2
Stanton's formula for a square r

[P n
= 5.13 sin aAV 2

plate becomes )

Stanton's formula for a plate

[P n =
7-70 sin aAV 2

twice as broad as long becomes )

The above formulae are not very consistent with each other.
The question of planes advancing at various angles through water
is in need of a complete and accurate experimental investigation.
It maybe noted that Stanton's plane twice as broad as long
approaches somewhat the proportions of an ordinary rudder of
barn-door type, and his coefficient for such a plate agrees well
with Joessel's results, which have been used a good deal for rudder
work in France. In England, the so-called Beaufoy's formula has
been much used for rudders. This gives P n = 3.2 sin a AV2 a ,

value much below

that from Joessel's formula. But in using this
formula, the center of pressure is assumed to be at the center of
figure instead of forward of it as by Joessel's formula for center
of pressure. The net result is that the English formula gives a
twisting moment on the rudder stock at usual helm angles only
about 30 per cent less than that derived from Joessel's complete
formulae. This is for ordinary rudders. For partially balanced
rudders the difference is somewhat less.

Experiments with rudders have indicated normal pressures on

them materially less than and sometimes but a fraction of what
would be given by Joessel's formula when V was taken as the

speed of the ship. But the true speed of a rudder through the
water in its vicinity is nearly always less and often much less than
the speed of the ship, and there are other conditions wherein a
rudder very much from a detached plate.

8. Eddy Resistance Formulae Applicable to Ships. All things

considered, it seems well, pending more complete experimental

investigation, to use for a plane Rayleigh's formula for front face
resistanceand Joessel's for total resistance.
Then we would have for a square stem, the end of a bow tor-

pedo tube, and similar fittings having head resistance only,

Pn '=2. 5 AV\
For square stern posts and similar objects P=r 2.1 AV 2
and for

scoops, square or nearly square to the surface of the ship, and

similar fittings, n
= 4.6 A V2 PIn these formulae A is area in

square feet, V is speed of the ship in knots and Pn , etc., are in

It is probable that these formulas would nearly always over-
estimate the resistance concerned, but as the resistances to which
they apply constitute a very small portion of the total in most
cases, it is not necessary to estimate them with great accuracy
and it is advisable to overestimate rather than underestimate
The resistance of struts is largely eddy resistance, but methods
for dealing with them will be considered in connection with ap-
9. Formula for Eddy Resistance behind Plate has Limitations.
In connection with the formula suggested for rear suction, namely
Pr = 2.1 2
it should be pointed out that this cannot apply as

speed is increased indefinitely.

Consider a plane of one square foot area immersed 10 feet say.
The pressure on its rear face, allowing 34 feet of water as the
equivalent of the atmospheric pressure and taking water as sea
water weighing 64 pounds per square foot, would be 44 X 64 =
2816 pounds. Evidently there is maximum rear suction when
there is a vacuum behind and no pressure on the rear face. Hence
2816 pounds is the maximum possible rear suction. By the for-

mula,if P= r 2816 = 2.1 F2 ,
F2 = = 1341, V = 36.62. Then

the formula obviously cannot apply beyond V = 36.62. Even if

the constant 2.1 is too great we will still in time reach a speed
where any formula of this type will give a rear suction equal to
the original forward pressure. Any formula which assumes that
suction increases indefinitely as the square of the speed must then
be regarded as expressing not a scientific fact but a convenient
semi-empirical approximation to the actual facts over the range of
speeds found in practice.

10. Wave Resistance

In discussing the disturbances of the water by a ship we have

given some consideration to the waves produced. To maintain
these waves, energy must be expended which can come only from
the ship. That portion of the ship's resistance which is absorbed
in raising and maintaining trains of waves is conveniently called
Wave Resistance.
i. Bow and Stern System. The tendency is toward the for-

mation of two distinct series of waves one initiated at the bow

and conveniently called the Bow Wave System and the other in-
itiated at the stern and called the Stern Wave System. The Stern
Wave System, however, makes its appearance in water already
more or less disturbed by the Bow Wave System and hence the
ultimate wave disturbance is compounded of the two systems.
When considering Kelvin's wave system as illustrated diagram-
ma tically in Figs. 60 and 61, we saw that it was made up of trans-
verse crests and diverging crests, the transverse crests being but
little curved and extending to the cusp line on each side. For
a given speed the length between successive transverse crests is
the same as the trochoidal wave length for the same speed.
It is evidently a reasonable approximation under the circum-
stances to substitute for the actual wave systems ideal systems
composed of traverse trochoidal waves extending out to the cusp
lines of Kelvin's waves and each wave of uniform height such
that energy of the ideal systems is the same as that of the actual


Consider system. To maintain this system there

first the bow
must be communicated to it while the ship advances the length
of one wave energy proportional to the energy of one wave

we denote by / the length from crest to crest of the tro-

choidal wave, by b its mean breath and by H its height, w being

the weight of water per cubic foot, we know from the trochoidal
wave formulae that the energy per wave length is proportional to
wblH 2 Now the external energy communicated to the system by

the wave resistance R w while the ship traverses a wave length /

is proportional to RJ. Hence RJ, is proportional to wblH2
or R w oc wbH 2
A similar formula applies to the stern wave

2. ResultantSystem. Wave
The actual wave resistance is
due to the wave system formed by compounding the bow and
sternwave systems. To determine the resultant system we com-
pound the bow and stern wave systems by the formulae for com-
pounding trochoidal waves.
In order to determine the resultant of the two separate wave
systems of the same length advancing in the same direction, we
need to know the distance between crests, and it is advisable to
consider the first crest of each system. The first crest of the
bow wave system will be somewhat bow and the first
abaft the
crest of the stern wave system somewhat abaft the stern. Their
positions and the distance between them will vary with speed.
Call the distance between them the wave-making length of the

ship and denote it by mL, where m is a coefficient varying slightly

with speed and, as we shall see, somewhat greater than unity.
Now, if V is the speed of the ship in knots, the bow wave length I
in feet is .5573 F
The distance between the first stern system crest

and the bow system crest next ahead of it is evidently the remainder
after subtracting from mL the lengths of the complete waves, if

any, between the first bow crest and the first stern crest. Let there
be n such waves and let the distance between the first stern sys-
tem crest and the bow system crest next forward of it be ql, where
/ is the wave length. Then mL = (n + q) I
= (n + q) .5573 F2 ,

where w is a whole number and q is a fraction. In the compound


wave formula we need know - or cos -

to cos Now, a in the
J\. I

u j j.i At 2 ira 2 irql

above is evidently the same as
ihen cos = cos 3-
r =

cos 2 irq. Now w being a whole number, cos 2 irq

= cos 2 TT (<7 + n)
mL Hence cos
2 ira
= cos 2 irmL 360 m
5573 V2 I -5573 V2 -5573 Z_
^2 U 2 T-U
Then cos
2 ^- = cos
m* <0
L by
c .
646 .
I c

The whole bow system is not superposed upon the stern system,
but only the inner portion, since the natural bow system extends
transversely to a greater distance than the natural stern system.
Let HI denote the height of the natural bow system when it has
spread to a given breath b, 2 the height of the natural stern H
system when it has spread to the same breath. Let kHi denote
the height of the natural bow system where the stern system has
spread to the breath b. Suppose its breath then is b'. Since it
has lost no energy bH* = b'k^H^.
Then the energy per wave length of the compound system re-

sulting from the superposition of a portion of the bow system of

breath b upon the whole stern system of breath b is measured by

Ib \&Hi*+ # + 2
2 #!# 2 cos 646 I.

The energy of the portion of the bow system beyond the stern
system and not compounded is measured by

/ (b'- b) &H? = l (b'&Hf- bVHfi = Ib (Hf- PHfi,

since b'k 2Hi 2 = bHi 2 .
Adding the above expressions for partial
energies the total energy per wave length is measured by

Ib ( H +# +
2 kHtHt cos 2 646).
\ c I

Whence the wave-making resistance is proportional to

(H?+ H 2 +
b 2 kHtHt cos ^ 646).

Now, b being an arbitrary convenient constant width, we can say

that the wave-making resistance R w is proportional to

#i 2 +# + kHA cos

3. General Formula Connecting Wave Resistance and Speed.

- The above wave
expression for resistance is of little quanti-
tative value without knowledge of coefficients appropriate to all
and for practical use in estimating wave resistance there are
methods more desirable than the use of a formula, but the expres-
sion is of value in enabling us to realize the general nature of the
wave resistance with speed.
variation of
As a step in this direction we need to know the connection
between HI and HI and the speed.
We know that in perfect stream motion the excess of pressure
near the bow is proportional to the square of the speed. If, then,
the wave height were proportional to the excess pressure, which
it must be approximately, since the surface pressure does not
change, we would have HI proportional to F Similarly H 2

would be proportional to F and we would have as the general


expression for R w the wave resistance,

R W = W^t+^BH- 2 kAB cos


The coefficients A and B are not constant. There are two

main sources of variation. If the bow wave height were always
proportional to the excess bow pressure as speed increases, A
would not vary on this account. It seems probable that at mod-
erate speeds when wave resistance first becomes of importance the
bow wave height does vary as the excess pressure, but as speed
increases a greater proportion of the stream line pressure is absorbed
in accelerating the water aft in stream line flow and a less pro-

portion in raising the water level. The same reasoning applies to

the stern wave, so, from this point of view, we would expect A
and B to be approximately constant at low and moderate speeds
and to fall off steadily at high speeds.
There is another important source of variation in A and B.
Suppose we have a vessel 400 feet long. Then the length of the

fore body is 200 feet. At 13 ^ knots the length of the bow wave
from crest to crest is very nearly 100 feet; at 27 knots it is 400
feet. at 13^ knots the bow wave is formed by the forward
quarter of the ship, as it were, while at 27 knots the whole for-
ward half of the ship must come into play. The result is, of
course, a modification of A and B
There appears to
with speed.
be a critical speed at which the wave length and the wave motion
and pressures are in step, as it were, with the ship, and the wave is
exaggerated. This may be called the speed of wave synchronism.
Broadly speaking, we may say that for fine models of cylindrical
below .55 the speed of wave synchronism in knots is
above \/Z, while for full-ended models of cylindrical coefficient
above .6 wave synchronism is below Vz. We may
the speed of
A and B as we approach the speed of
expect to find a rapid rise of
wave synchronism and a less rapid falling off as we pass beyond it.
Consider now the coefficient k in the formula

Rw = F 4
+ B + 2 kAB
cos ^
\ c

At low speeds k is evidently zero, since observation shows that at

low speeds the bow disturbance has spread out abreast the stern
to a distance where it is not affected one way or the other by the
stern disturbance. As the speed increases, however, more and
more of the bow wave energy is found in the vicinity of the stern
and k may be expected to become greater and greater. It is also
a matter of observation that for narrow deep models the trans-
verse features of the bow wave are accentuated, and hence for
such models k will, other things being equal, be greater than for
broad shallow models, since it is the transverse portion of the bow
system which is available for combination with the stern system.
Consider, now, finally, the term cos 646. This expression is

equal to +i when 646= 360 or any multiple of 360. It is

equal to i when 646= 180 or 180+ any multiple of 360.
The quantity m is approximately constant for a given ship, though
it increases somewhat with the speed. It also appears to increase

somewhat with fullness from ship to ship. A fair average value of

m would seem to be about 1.15 for speeds where humps and hol-
lows are of importance. For lower speeds m approaches i. Fig.
65 shows for m = 1.15 a curve of cos 646 plotted upon c or

It is seen that at low speeds maxima and minima succeed one
" "
another very rapidly. Each maximum corresponds to a hump
in the curve of residuary resistance and each minimum to a
Humps and hollows on actual resistance curves do not manifest
themselves, however, in accordance with Fig. 65. The varying
term is kA B cos 646, and since in most cases at low speeds k

is so small as to be practically negligible, we find in practice that

the first important hump usually appears for full models at about

= = i ,
while for fine models this hump is imperceptible or shows
itself only as an unfair portion of the curve and the first important
V =
hump is at about 1.4 to 1.5.

For quite full models, especially those with parallel middle

V =
body, the hump for .8 is often important, and for such models
y- =
the hump for .67 to .7 is frequently detected though not of
The values of = above refer to the centers of the humps or the
points where the percentage increase of resistance above an aver-
age curve is a maximum. Of course, the departures from the
average begin and end some distance before and beyond the hump

Fig. 66 shows graphically the relations between speed of ship,

length in feet and values of = By using a varying scale for

length, the abscissae being proportional to \/length, the contours

of =r are straight lines. By shading the regions corresponding to

humps and leaving clear those corresponding to hollows the rela-

tive locations of humps and hollows are indicated. It will be
observed that the two lower humps of Fig. 66 are indicated at
slightly lower values of = than in Fig. 65. This is because Fig. 65

is for a constant value of m, namely 1.15, while in practice we find

for the lowest hump m = i.oo very nearly, and for the next
V V =
m = i. 08 or so. For the region from -= = .9 to = 1.2, embrac-
ing a hump and a hollow, m= 1.15 very nearly while beyond this

speed m
somewhat greater on the average.

It might seem at first sight very important to adopt such length

be sure of landing in a hollow rather than
for a desired speed as to
on a hump, but, though this point should always be considered, in
comparatively few cases is it a matter of serious practical impor-
tance. In most cases it is desirable to adopt proportions and form
such that the humps and hollows up to the speed attained are not
prominent, so there is no material saving to be had by landing in
a hollow rather than on a hump.
4. Curves of Residuary Resistance and of Coefficients. Hav-
ing discussed generally the characteristics of wave-making re-

sistance as indicated by the formula

Rw = F
^ +5 + kAB
4 2 2
2 cos
it is well to consider some concrete examples.
Fig. 67 shows curves of residuary resistance determined from
model experiments for ten 4oo-foot ships without appendages.
The residuary resistance
is practically all wave-making. The
proportions, etc., are tabulated on the figure.
It is seen that there are five displacements in all, there being
two vessels of each displacement differing in midship area or lon-
gitudinal coefficient. All vessels were derived originally from the
same parent lines, so the variations of resistance are essentially
due to variations of dimensions and of longitudinal coefficient.

The curves of Fig. 67 are not very encouraging to the develop-

ment of an approximate formula for wave resistance.
For instance the variation with longitudinal coefficient is a
very difficult feature. The models of .64 longitudinal coefficient
all show pronounced humps at about 21 knots, while their mates

of .56 longitudinal coefficient show no hump there. But at 25

knots or so the wave resistances for the two coefficients come
together again, and for higher speeds the models of .64 coefficient
have the smaller resistances. At 30 knots or so there is a second
hump which shows for both the full and the fine coefficients.
Resistance curves are frequently analyzed by assuming them of
the form R = n
and determining suitable values of n, the
power which the resistance is varying,
of the speed according to
and of a, the corresponding coefficient. The curves of Fig. 67 are
analyzed in this way without much trouble by plotting them upon
logarithmic section paper. For a curve so plotted the exponent n
at a point is proportional to the inclination of the curve.
Fig. 68 shows curves of the exponent n for the 10 curves of
wave resistance of Fig. 67. It is seen that the variations of n
are enormous. As to a in the formula R = aVn the values cor-

responding to the curves of n in Fig. 68 vary too rapidly and radi-

cally to be adequately represented graphically.
Suppose now we attempt a slightly different analysis. We have
deduced a qualitative formula for wave resistance as follows:

m6 \ , ,

? 46j.
Then curves of -
will also be curves of

2 kAB cos 646


and might be expected not to vary very much. Fig. 69 shows

curves of for the 10 curves of Fig. 67, the residuary resistance

R being taken as identical with R w

It is seen that up to 18

knots or so these curves are reasonably constant. Here they

begin to rise. For the full coefficients there is a maximum at 21

knots or so, a minimum at 23 to 24 knots and a second maximum

at 29 to 30 knots. For the fine coefficients there is only one pro-
nounced maximum at 29 to 30 knots.
It is evident from Fig. 69 that the curves of ~ are somewhat

systematic in their variations and that it might be possible to for-

mulate values of A, B, k and m
such that in a given case we could
determine Rw with reasonable approximation from the basic for-

Rw = F 4
+ B + 2 kAB
\ C I

It is and c involved
equally evident that the formulae for A, B
would be difficult and complicated. It will be shown later that by
graphic methods the residuary resistance in a given case can be
readily approximated and hence the task of devising approximate
formulae need not be undertaken.
It is interesting to note for ships i to 4 the relative reduction in
wave resistance beyond 30 knots.
The reason will be made clear upon reference to Fig. 66. It is
seen that for a 4oo-foot ship the last hump occurs at about 30
knots. In this condition the wave length corresponding to the
speed is somewhat greater than the length of the ship, so that the
second crest of the bow wave is superposed upon the first crest of
the stern wave. Hence the hump. At a speed of about 40 knots
there would be a final hollow corresponding to the conditions
when the first hollow of the bow wave is superposed upon the
first crest of the stern wave. This is the main cause of the ap-

parent relative falling off of wave and 69

resistance in Figs. 68
between 30 and 40 knots.
Fig. 66 would indicate that some distance beyond 40 knots the
wave resistance of these 4OO-foot ships would again begin to in-
crease relatively, but there is some reason to believe that at excessive

speeds say 120 knots for the 400-foot ships the wave resist-
ance would be decreased by the bodily rise of the ship, which
would begin to approach the condition of a skipping stone and
tend to glide along the surface. Of course, the speed of 120 knots

isunattainable by any 4oo-foot ship at present, but it corresponds

to 36 knots for a 36-foot boat, which is not very far beyond the

speed-launch results now attainable. Consideration of such ex-

treme cases is, however, beyond the scope of this work.

ii. Air Resistance

The above water portions of a ship may be regarded as im-

mersed in the air, and air, like water, offers resistance to the motion
of a body surrounded by it. Air is, roughly, only from one-ninth
to one-eighth of one per cent of the weight of water, the actual

weight depending on the pressure and temperature, and air re-

sistances compared with those of water are, roughly, as the rela-
tive densities. But air resistance is by no means always negli-
gible. Sailing vessels are driven by the resistance of sails to the
motion of air past them, and any one who has attempted to stand
on the deck of a vessel exposed to a gale of wind will admit that a
strong head wind opposes a good deal of resistance to a vessel with
even a moderate amount of top-hamper.
i. Zahm's Experiments upon Air Friction. Air resistance can
be separated into two classes frictional and eddy resistance.

Careful investigations of the friction of air upon plane surfaces

have been made by Prof. A. F. Zahm, of Washington, who in a

paper of February 27, 1904, before the Philosophical Society of

Washington (Bulletin, Vol. XIV,
pp. 247-276) has given experi-
mental results for air friction upon thin planes somewhat similar
to those tried in water by Froude.
Prof. Zahm's air planes were 25? inches wide, one inch thick, and
of varying lengths up to 16 feet. While rather smaller than
Froude's planes, they were tried up to a high air velocity of 25
statute miles per hour, or 2if knots.
Prof. Zahm summarizes his most important conclusions upon
the subject of air resistance as follows:
1. The total resistance of all bodies of fixed size, shape and
aspect is expressed by an equation of the form R
= av R being n

the resistance, v the wind speed, a and n numerical constants.

2. For smooth planes of constant length and variable speed,,

the tangential resistance may be written R = fa;

1 85 '


3. For smooth planes of variable length / and constant width

and speed the friction is R = c/' 93 .

4. All even surfaces have approximately the same coefficient of

skin friction.

5. Uneven surfaces have a greater coefficient of skin friction,

and the resistance increases approximately as the square of the

These conclusions as to air friction are in striking agreement
with those deduced by Froude for surface friction in water.
The coefficients given by Zahm are readily reduced for speeds
in knots instead of feet per second or statute miles per hour.

Upon doing this, if R denote frictional air resistance in pounds,

A denote whole area of surface in square feet, / denote length of
surface in feet and V denote speed through the air in knots, we

have R= .0000122 I'

AF 1 '

It should be remembered that this formula is based upon ex-

periments with planes no longer than 16 feet tested up to speeds
of 25 statute miles per hour. So, while it may be used with con-
fidence for short planes up to any velocity reached by ships, it
must be regarded as only a fair approximation for long surfaces.
Fortunately for the purpose of the naval architect a fair approxi-
mation to frictional air resistance is all that he ever need know in
practice. It is very seldom indeed that he will need to take any
account of it at all.

For convenience in calculation Table VIII gives values of F1 ' 85

and of r^- We have /'

93 = and hence can readily obtain /

we know A table of l'

would not admit of easy interpolation,

while ' which varies comparatively slowly, lends itself to inter-


Comparing the results of his experiments on air friction with
those of Froude on water friction, Zahm states:
With a varnished board 2 feet long, moving 10 feet a second,
the ratio of our coefficients of friction for air and water is 1.08

times the ratio of the densities of those media under the con-
ditions of the experiments."

Froude, however, found that the coefficient of friction fell off

more rapidly with length than as /~' 07 so that for longer planes the

above ratio is greater than i .08 times the ratio of densities. Thus,
for 2o-foot planes the ratio of coefficients would be some ii times
the density ratio, that is, the friction in air would be i? times that
deduced from water friction by dividing it by the density ratio.
Zahm states that in his experiments " no effort was made to
determine the relation between the density and skin friction of
the air, partly for want of time, partly because, with the apparatus
in hand, too great changes of density would be needed to reveal
such relation accurately. Doubtless the friction increases with
the density."
It appears probable that we may assume Zahm's formula for
frictional resistance of air to apply to air at 60 F. and a barome-
ter pressure of 30 inches.
2. Eddy Resistance in Air. Results of Experiments with
Planes. While the frictional resistance of air is of
importance in
connection with flying machines, for ships the most important air
resistance is the eddy resistance.
The eddy same general laws
resistance of air seems to follow the
as the eddy resistance of water. Within the limits of the speed
attained by the wind, say up to 100 miles per hour, it varies for a
given plane as the square of the speed. Observations made under
the direction of Sir Benjamin Baker during the construction of
the Forth Bridge indicated that small planes exposed to the
wind offered greater resistance per square foot than larger planes
exposed to the same wind. M. Eiffel found for planes not over
i meter square falling through still air that the larger planes showed

slightly greater resistance per square foot.

For rectangular planes the resistance varies somewhat with the
narrow plane offering greater resistance
ratio of the sides, a long
than a square of the same area.
For our purposes it is not necessary to consider closely these
minutiae, and it will suffice to use an average coefficient and ex-

press the resistance in pounds of a plane of area A in square feet


moving normally through the air with velocity V knots by a single

formula 7? = CAV\
The values of the coefficient C which have been obtained by
various experimenters vary a good deal. The more recent experi-
menters seem to obtain the lower values, but coefficients obtained
by experimenters within the last 30 to 40 years range from .0035
to .005 about.
In England, Stanton, with very small planes exposed to a cur-
rent of air through a large pipe or box, has obtained a coefficient
of .0036. Dines with rather small planes on a whirling arm has
obtained .00384. Mr. William Froude with good-sized planes mov-
ing through still low velocities obtained .0048. In
air at rather

America, Langley, by whirling-arm methods, obtained somewhat

variable coefficients averaging about .0047. In France, quite
recently, M. with planes up to 10 square feet or more in

area, falling through still air, conducted very careful and elab-
orate experiments and obtained a coefficient of .004. (See Re-
cherches Experimental sur la Resistance de PAir Executees a la
Tour Eiffel This was published in 1907.)
par G. Eiffel."
All things considered, in the light of our present experimental

knowledge on the subject appears reasonable to adopt the


coefficient .0043 as suitable for practical use. Then our formula

for the resistance in pounds of a plane moving normally to itself is
R = .0043 A V where A is area in square feet and V is speed in

knots. For speed in statute miles the coefficient above should be

divided by 1.326; for speed in feet per second by 2.853.
When it comes to the normal pressure on an inclined plane
moving through the air the results obtained by experimenters are
somewhat peculiar. For square planes and rectangular planes
whose sides are not too dissimilar the normal pressure increases
rapidly from zero at zero inclination up to an inclination of 30
degrees or so. At this inclination the normal pressure is nearly
the same as at 90 inclination, and from 30 to 90 inclination the
normal pressure, while varying somewhat irregularly, does not
change much.
The simplest formula is that of M. Eiffel. For inclined planes
he proposes to take the normal pressure as constant from 30

to 90, and from o to 30 to take it as varying linearly. The

Eiffel formula
a sufficiently close approximation for practical use.

The formula, then, for practical use expressing the normal pres-
sure in pounds P n on an inclined plane moving through the air at
an angle of degrees will be
From o to 30, Pn = .0043 AV 2

above 30, Pn = .0043 A V 2

where A is area in square feet
and V is speed in knots.
The normal pressure is, of course, different from the resistance
in the direction of motion, which is Pn sin 6, or the component of
Pn parallel to the direction of motion.
3. Determination of Air Resistance of Ships. There is no prac-
ticalmethod recognized at present for determining the air resist-
ance of a ship. Mr. William Froude made some experimental
investigations of the matter about 1874, in connection with the
Greyhound, a vessel 172.5 feet X 32.2 feet X
13 feet draught, of
about 1000 tons displacement. The vessel was tried without masts
or rigging. He concluded that in this condition at 10 knots, the
air resistance of the Greyhound was nearly 150 pounds, or about
i per cent of the water resistance.
For steamers without large upper works, the air resistance,
when the air is still, is, without doubt, too small as a rule to re-
quire much consideration. With a strong head wind the air
resistance is, of course, very much increased, but under such con-
ditions the increase of water resistance due to the head sea is

probably in most cases far greater than the air resistance. In

cases where air resistance is important, it can be investigated by
exposing a model with the upper works complete to a current of
air of known speed. The law of the square applies, and it will be
possible to determine the air resistance of the model at the actual
speed, not the corresponding speed of the ship. Then the air
resistance of the full-sized ship, being practically all eddy re-
sistance, may be estimated by multiplying the resistance of the
model at the speed of the ship by the square of the ratio between
the linear dimensions of the ship and the model.

For a rough approximation, we may take the area of the por-

tion of the ship above water projected on a thwartship plane and
assume that the air resistance is that due to a plane of this area
denoted by A advancing normally through the air, using the
formula already given for the resistance of a plane. This would
give us
Air Resistance in pounds = .0043 A V2 where V is speed through

the air in knots and A is area of upper works projected on a thwart-

ship plane.

12. Model Experiment Methods

In view of the very large use now made of model-basin experi-
ments there will be given a brief description of the methods used
in deducing from the model experimental results the resistance or
effective horse-power of the full-sized ship.
At a model tank or basin there are facilities for making to scale
models of ships representing accurately the under-water hulls and
a sufficient amount of the above-water hulls. Most model basins
work with models from 10 Some use models as
to 12 feet long.

long as 20 feet. A
complete model can be towed through the still
water of the basin, the speed and corresponding resistance being
measured for a number of speeds covering the range desired.
i. Treatment of Model Results. By plotting each resistance
as an ordinate above its speed as an abscissa we obtain a number
of spots through which a fair average curve is drawn, giving the
total resistance of the model. Fig. 70 shows for an actual model
a number of experimental spots and the resistance curve drawn
through them. When reducing the results the first step in practi-
cally all cases is to determine the estimated frictional resistance
of the model.
The wetted surface of the model has been calculated and we
have recorded from experiments with planes the length of the
model, the resistance of a square foot of surface for each tenth of a
knot extending up to any speed to which a model is likely to be
tested. Fig. 70 shows a curve of rf or frictional resistance of
model, its ordinates having been determined for various speeds by

multiplying the model surface by the resistance of one square foot.


2. Deduction of Ship Resistance, Using Model Results. For

the most common case the model represents some full-sized ship
- actual or and we wish by the aid of the model
results to determine a curve of estimated effective horse-power
for the full-sized ship.
Table IX herewith gives the calculations for the Yorktown, for
whose model the resistance curve is given in Fig. 70. The object
of much of the form is obvious. The Mean Immersed Length,"
L, of the ship is usually the length on the load water line. For
models of peculiar profiles there is a correction applied by judg-
ment, the object being to obtain the average immersed length.
The mean immersed length of the model is usually made 20 feet
at the United States Model Basin, though moderate departures
from this length are made when desirable for any reason. Also,
as it is difficult to get satisfactory observationsabove a speed of
17 knots of model, it is necessary to make models
shorter than 20
feet if the maximum corresponding speed would be over 17 knots
for a 20-foot model.
The model is so weighted that if it is exact it will float in the
fresh water of the basin at exactly the corresponding water line of
the ship in salt water. Hence the ratio at corresponding speeds

/ZA 3 ^6 (L\ 3
of resistances which follow Froude's La.w is not :
but a- ( )i the

factor *7 being introduced on account of the passage from fresh
water to salt water.
Coming now to the tabular form, there are entered in the first
column values of v or the speed of the model in knots, and in the
second column corresponding values of r or the total resistance of
the model in pounds as taken from the curve in Fig. 70. In the
third column is entered rf or the frictional resistance of the model
calculated as already described. In the fourth column we
enter the residuary resistance, r r which is equal to r
r{ It is .

this resistance to which Froude's Law applies, and we wish to de-

duce from it in the shortest and simplest manner the correspond-
ing residuary effective horse-power. While r r is mostly Wave
Resistance, it includes the Eddy Resistance and Air Resistance

of the model. Both are taken as following the Law of Com-

Now for the full-sized ship the residuary resistance in pounds at
^6 /ZA =
corresponding speed is rr X (y) -^r say. The speed of ship,

V, corresponding to a speed of model, v, is v

l/y, and the effective

horse-power absorbed by R r is R X r .0030707 V. Then, if the

residuary effective horse-power for the full-sized ship is denoted
by EHP r we have
= R X .0030707 V = 36 /ZA I~L
EHP r r rr >

y J X .0030707
( v t/
35 \t / I

36 /ZA /Z
We denote by a the quantity
(y) .0030707 Uy and calculate it
35 \ / /

once for all, as indicated in the heading. Then in the fifth column
of the table we enter av and in the sixth column r which is EHP ,

simply rr multiplied in each case by av. In the ninth column we

enter V, the corresponding speed for the ship, obtained by multi-

plying each value of v by V/y We have now for a number of

values of V the values of EHP r or residuary effective horse-power.

We need to determine the frictional portion of the effective
horse-power. denoted by Ef or
This is f To determine EHP .

frictional resistance we take from Table VI of Tideman's Con-

stants the coefficient of friction appropriate to the length of the
vessel and the nature of bottom. The area of wetted surface
has been calculated.
We have seen that frictional resistance in pounds = Rf =
wetted surface X frictional coefficient X V 1 83'

and E/= .0030707 Rf X V

= .0030707 X wetted surface X frictional coefficient X F2 '

Taking from Table VII the values of F2 83 we


readily determine
and enter in column n the values of EHPf . These values are
plotted as in Fig. 71 and a fair curve run through. Then from

this curve for the values of V cor in column 9 we take off the
values of EHPf and enter them in column 7. Column 8, which
is the sum of columns 6 and 7, gives the values of the total EHP,
which, spotted in Fig. 71 over the values of F cor enables us to ,

draw the final curve of E.H.P. for the condition of the ship defined
in the heading of the table.
From this curve it is possible to fill in column 12, which gives
the values of E.H.P. corresponding to the even values of V in
column 10. Column 12 is, however, seldom needed.
Residuary Resistance Plotted for Analysis.
3. When we are
dealing with an actual ship or design it is generally desirable to
deduce from the model results the final E.H.P. curve as soon as
possible. When, however, it is a question of analysis of residu-
ary resistance it is desirable to express it in a slightly different
form. A very convenient and instructive method is to use the
values of as abscissae and of Resistance -H Displacement as
For convenience the value of Resistance H- Displacement is

expressed as Resistance in Pounds per Ton of Displacement.

Fig. 72 shows the curve of Residuary Resistance in Pounds per
Ton plotted on V +vL
for the model to which Figs. 70 and 71

refer. Fig. 72 is applicable to any size, and it is this elimination of

the size feature which renders this method of plotting of value
for purposes of analysis.

13. Factors Affecting Resistance

The problem of resistance in its most general form involves too

many variables to be capable of experimental solution. For a

vessel of given displacement and speed the resistance varies with
variations of (i) The dimensions, (2) The shapes of water lines
and sections. For a vessel of given displacement we may have
an infinite number of variations of dimensions and shape, so even
if we could deduce the resistance of a vessel with mathematical
accuracy from model experiments, it would be a formidable under-
taking to investigate all admissible or likely variations of dimen-
sions and shape for but a single vessel of a fixed displacement.

I. Derivation of Models from Parent Lines. If, however, we

adopt a single definite shape or set of parent lines, deducing all
models from these lines by variations of dimensions and coeffi-
cients of fineness, the problem is enormously simplified. By test-
ing a practicable number of models we can determine, not for
one displacement only, but for any displacement within a certain
range and for any dimensions and fineness likely in practice, the
approximate resistance at any practicable speed.
In connection with fineness the expression longitudinal coeffi-
cient" will be used to denote the ratio between the volume of dis-
placement of a vessel and the volume of a cylinder of section the
same as the submerged midship section and of length the same as
the length of the vessel preferably the mean immersed length.
This coefficient sometimes called the "cylindrical coefficient" and

very commonly the "prismatic coefficient." While cylindrical coef-

ficient is descriptive and correct, it is thought that the designation

''longitudinal coefficient" is preferable as emphasizing the fact that

this coefficient measures and expresses the fineness of the vessel in

a longitudinal direction. The
expression "prismatic coefficient" is
slightly in error, since strictly speaking the section of any prism
is bounded by a straight-sided polygon and not by a curve.

Given a set of parent lines, the deduction from them of lines of the

same coefficients but of different proportions or relative values of

length,beam and draught, is a simple matter. If length alone is
changed, we need only change the spacing of stations in propor-
tion to the change of length. If draught alone is changed, we

need change only in a corresponding way the spacing of water

lines. If beam alone is changed, we need change only the ordi-
nates of water lines.

Since the changes caused by change of length, beam and draught

are independent we may simultaneously change all three, if we
wish, without difficulty.
Suppose, however, we wish to keep dimensions unchanged and
make changes shape and fullness. We cannot change the

midship section without departing from the parent lines, but we

can change in a comparatively simple manner the longitudinal
coefficient or curve of sectional areas. Thus in Fig. 73, suppose

the curve numbered i is the curve of sectional areas for the parent
model and the curve numbered 2 the desired curve of sectional

areas. Through ,
the point on curve 2 corresponding to the
AB, draw EF horizontally to meet curve i at F. Through
F draw CD, then the proper section at AB of the derived form is
the section at CD of the parent form.
Having the two curves of sec-
tional area and the half -breadth plan of the parent form, any desired
section of the derived form can be determined without difficulty.
From a single parent form then, we can derive forms covering
all needed variations of displacement, of proportions and of fine-

ness as expressed by "longitudinal coefficient." By contour curves

from the results of a number of models derived from one parent
form we can deduce diagrams enabling us to ascertain the resist-
ance at any speed of any vessel upon the lines of the parent form.
This applies, of course, to residuary resistance only, since the fric-
tional resistance can always be estimated without model results.

or experiments in the manner already indicated.

2. Classification of Factors Affecting Resistance. It would
require experiments with models derived from an infinite num-
ber of parent forms to trace the effect of all possible variations of
shape, but if we can determine the major factors affecting resist-
ance and their approximate effect we need seldom concern ourselves
with the minor factors.
While necessary to be cautious in laying down from past
it is

experience a hard and fast line of demarcation between the major

and minor factors of resistance, since novel developments in the
future may convert one into the other, yet so far as can be judged
from trials at the United States Model Basin of over a thousand
models we appear warranted in drawing some conclusions as ta
the principal factors affecting the resistance of ships not of abnor-
mal form and the relative importance of these factors. We need
consider only frictional and wave-making or residuary resistance.
Given the displacement, speed and frictional quality of the
surface, the only other factor of importance as regards frictional
resistance isthe length. The greater the length for a given dis-

placement the greater the frictional resistance. This because

frictional resistance is proportional to surface or \^DL.

As regards residuary resistance for a given displacement, the

principal factors arranged in their usual order of importance are

as follows :

1. The length.
2. The area of midship section or, conversely, the longitudinal

3. The ratio between beam and draught.

4. The shape of midship section or midship section coefficient.

5. The details of shape toward the extremities.

It is seen that factors i, 2 and 3 can be investigated from a
single parent form. The complete investigation of factors 4 and 5
would require investigations involving a very large number of

parent forms. Fortunately, however, these factors are those of

least importance.

3. Details of Shape Forward and Aft. In placing factor 5 as

of small importance, it should be understood that this is the case

only as regards the variations found in good practice. If abnormal

shapes for the extremities are adopted, abnormal resistance is
liable to follow. The dictum of William Froude many years ago
appears to be still our best guide. He stated that, broadly speak-

ing, it was desirable to make the bow sections of U

shape and the
stern sections of V
shape. This amounts to saying that at the
bow it is^ advisable to put
the displacement well below water and
make the water line narrow, and at the stern it is advisable to
bring the displacement up towards the surface and make the
water line broad. Carried to an extreme, this would give us
hollow water lines at the bow and the broad flat stern of the
torpedo boat type. As a matter of fact, model basin experi-
ments appear to indicate that for smooth water, up to quite a
high speed, this type of model is about the fastest. For extreme
speeds, even in smooth water, hollow bowlines are seldom
adopted, but there not sufficient experience in this connection

to say positively that they are or are not desirable from the point
of view of speed alone.
In this connection itbe pointed out that experiments show
a ram bow of bulbous type to be favorable to speed, even apart
from the fact that the ram bow usually involves a slight increase

in effective length. This issimply because the ram bow, which is

the extreme case of the U bow, is much fuller below water than at
the water line.

The excess pressures set up around the ram being well below
the surface are more absorbed in pumping the water aft, where it
is needed, and less absorbed in raising the surface and producing

waves than if the same displacement were brought close to the

There appears to be a reasonable explanation of the advantages
as regards resistance of the broad flat stern. In wake of the
center of length, the water is flowing aft to fill up the space being
leftby the stern, the greatest velocity of the water being under
the bottom. As the vessel passes, the water flows aft and up,
losing velocity all the while and increasing in pressure.

With a U stern there is little to check the upward component of

the velocity which absorbed in raising a wave aft. With the

broad flat stern against which the water impinges, as it were, more
or less of the upward velocity is absorbed by pressure against the
stern, which have a forward component, the result being a

closer approach perfect stream motion and less wave dis-

While the broad flat stern is slightly superior as regards resid-
uary resistance in smooth water, it is apt to have unnecessary
wetted surface and is objectionable from a structural and sea-
going point of view. With model basin facilities it is generally
possible to determine upon a stern of V type which is almost as
good as the broad flat type as regards resistance, and distinctly
preferable to it from a structural and sea-going point of view.
In connection with the details of shape forward and aft the
change of trim upon resistance may be considered, since
effect of
the principal effect of change of trim is to modify the shape
towards the extremities.
Any change of trim, no matter how small, necessarily pro-
duces some effect upon resistance, and there are many sea-going
people who ascribe great virtue to some particular trim and great
influence upon resistance to change of trim, generally considering
trim by the stern as advantageous for speed.

Trim by the stern has some advantages in that it generally

improves the steering of the ship or its steadiness on a course, and
in rough weather it is generally advantageous to secure greater
immersion of the screws and more freeboard forward; but as
regards resistance in smooth water changes of trim occurring in
practice generally produce changes of resistance of little or no
In 1871 Mr. William Froude investigated the effect of trim upon
the resistance of the Greyhound, a vessel 172 feet long and towed
at displacements from 938 to 1161 tons and at trims varying
from 1.5 feet by the head to 4.5 feet by the stern. The maxi-
mum speed at which the vessel was towed was about 12 knots.
These experiments showed that for the Greyhound trim by the
head was beneficial at low speeds, below 8 knots, and trim by the
stern was upper speeds, above 9 knots. The
beneficial at the

however, were comparatively small for quite large


changes of trim. Mr. Froude's conclusion from these full-sized

towing experiments was, As dependent on differences of trim, the
resistance does not change largely; indeed, at speeds between 8
and 10 knots it scarcely changes appreciably, even under the maxi-
mum differences of trim." The results from the Greyhound
were corroborated by model experiments which agreed quite
well with the full-sized results, and since these classical experi-
ments of Mr. Froude, model experiments investigating this ques-
tion have been repeatedly made.

Many experiments made at the United States Model Basin

appear to indicate that, broadly speaking, for the majority of
actual vessels at full speed a slight trim by the stern is beneficial,
but that in the vast majority of cases the benefit is too small to
be of practical importance. With a well-balanced design, the
fineness forward and aft being properly distributed, the effect

upon resistance of change of trim is practically nil.

4. Shape of Midship Section. Let us now consider the in-

fluence upon resistance of midship section fullness or the midship
section coefficient. Figs. 50 to 54 show body plans of five models,
all having the same length, the same displacement 3000 pounds
the same curve of sectional areas, the same area of midship

sectionand practically the same load water line. Figs. 55 to 59

show similarly body plans of five 1000 pound models.
Each group of five models has midship section coefficients vary-
ing from .7 to i.i, the models with fine midship section coefficients
having greater values of B and H
since the actual midship sec-
tion areas are the same for all models of a group. The ratio
B -f- H for all ten models is 2.92. The models are of moder-

ately fine type, the longitudinal coefficient being .56 for all ten.
Fig. 74 shows curves of residuary resistance in pounds per ton
for the five 3000 pound models and Fig. 75 shows similarly the
resistances of the five 1000 pound models.
It is seen that while the models with full midship section
coefficients drive a little easier up to F-S-vZ = i.i to 1.2 and
the models with fine coefficients have a shade the best of it at

higher speeds, the differences for such variations of fullness as

are found in practice are remarkably small. The results given
above are taken from a paper by the author before the Society of
Naval Architects and Marine Engineers in November, 1908, on
TheInfluence of Midship Section Shape upon the Resistance of

Ships." This paper contained many other results similar to those

given, and its conclusion was that for vessels of usual types and
of speeds in knots no greater than twice the square root of the

length in feet, the naval architect may vary widely midship section
fullness without material beneficial or prejudicial effect upon speed."
Of course, it follows that the minor variations in shape of midship
section that can be made in practice without changing fullness have
practically no effect upon resistance.
should be most carefully borne in mind that the above

applies to the shape and coefficient of a midship section of a given

area, not to the area of the section.
5. Ratio between Beam and Draught. Consider now the
effect of the ratiobetween beam and draught. Figure 76 shows
curves of E.H.P. as determined by model experiment for 6 vessels,
all derived from the lines of the U. S. S. Yorktown but varying in

proportions of beam and draught from a very broad shallow model

to a very narrow deep one.
It is seen that the broader and shallower the model the greater

the resistance. is typical and confirmed by many

This result
other experiments at the United States Model Basin. It may at

first sight seem opposed to many cases of experience where beamy

models proved easy to drive. But in these cases it will be found

that the increase of beam carried with it increase of area of mid-
ship section. Had beam been increased and draught decreased
in proportion, the area of midship section remaining unchanged,
the results would have been different.
However, the variations of resistance with variations of the ratio
of beam to draught are not very great as a rule.

Longitudinal Coefficient or Midship Section Area.

6. Take
up now the effect upon resistance of the variation of midsnip sec-
tion area or longitudinal coefficient. This is a factor of prime
importance in some cases and quite secondary in others. Thus,
Fig. 67 shows curves of residuary resistance for five pairs of 400-
foot ships, each pair having the same displacement and derived
from the same parent lines but differing in midship section area or
longitudinal coefficient. It is seen that at 21 knots No. 10 with
.64 longitudinal coefficient has 2.3 times the residuary resistance
of itsmate No. 9 with .56 longitudinal coefficient. But at 24^
knots they have the same residuary resistance.
Again, No. 4 of .64 coefficient at 21 knots has nearly twice the
residuary resistance of No. 3 of .56 coefficient. At 255 knots they
have the same residuary resistance and at higher speeds No. 4
has the best of it, having but .9 of the residuary resistance of No. 3
at 35 knots. These results, which are thoroughly typical, are sus-

ceptible of a very simple qualitative explanation. small longi- A

tudinal coefficient means large area of midship section and fine
ends. A large longitudinal coefficient means small area of mid-
ship section and full ends. At moderate speed the ends do the
bulk of the wave making and the fine ends make much less wave
disturbance than the full ends. Hence the enormous advantage of
the fine ends at 21 knots in Fig. 67. But at high speeds the whole
body of the ship takes part in the wave making and the smaller
the midship section the less the wave making. It follows that for
a ship of given dimensions, displacement, type of form and speed
there is an optimum longitudinal coefficient or area of midship

section. Data will be given later by which this can be deter-

mined with close approximation.
7. Effect of Length. There remains finally to .consider the
factor which, broadly speaking, has more influence upon residuary
resistance than any other. This is the length. We have seen
that for a given displacement the greater the length the greater
the frictional resistance it varying as L. V '

Residuary resist-
ance, on the contrary, always falls off as length increases, though
not according to any simple law. Fig. 77 shows curves or residu-
ary resistance of five vessels, all of 5120 tons, derived from the
same parent lines and having the lengths given. Of course the
longer Vessels have beam and draught decreased in the same ratio
sufficiently tokeep the displacement constant. Fig. 77 illustrates
very clearly the enormous influence of length upon residuary
resistance. Since frictional resistance increases and residuary
resistance decreases with length, it is reasonable to suppose that
for a given displacement and speed there will be a length for which
the total resistance will be a minimum. There is such a length,
but in the vicinity of the minimum the increase of resistance with
decrease of length is slow, and since length in a ship is usually
undesirable from every point of view except that of speed, ships
should be made of less length than the length for minimum resist-
ance. For men-of-war particularly
it is good policy to shorten the

ship, put in slightly heavier machinery and accept the increased

coal consumption upon the rare occasions when steaming at full

speed, rather than to lengthen the ship, carry greater weight of

hull and armor necessitated thereby, and consume more coal at

ordinary cruising speeds.

14. Practical Coefficients and Constants for Ship Resistance

Primary Variables Used.

i. The first thing to do when we
wish to establish methods for the determination of ship resist-
ance is primary variables to be used. In a given case
to fix the
we may have dimensions, displacement, etc., all fixed, and need to
determine the resistance at a given speed, or we may wish to de-
termine dimensions to bring resistance below a certain amount, or
the problem may present other aspects. The primary variables

adopted should enable the data available to be applied simply

and directly to the problems arising.
It is convenient to express resistance as a fraction of displace-

ment, and a suitable measure is the resistance in pounds per ton

of displacement. Then a resistance of one pound per ton of dis-
placement means a resistance which is uu ^
of the displacement.
At corresponding speeds for similar models, resistances which
follow Froude's Law
are proportional to displacement, and hence
the pounds per ton are constant.
Speed is conveniently expressed not directly but in terms of >

the speed length ratio or speed length coefficient. For similar

models at corresponding speeds is constant.

When it comes to size we need a variable which does not change

for similar models whatever the displacement. Since the dis-

placement varies as the cube of linear dimensions, such a quan-

tity would be Displacement -j- (any quantity proportional to the
cube of linear dimensions). As length is much more important
in connection with resistance than beam or draught, a suitable

quantity would be This would usually be a very small frac-


tion, however, and it is desirable to use a function which in prac-

tical cases assumes numerical values convenient for consideration

and comparison. Such a function is >

called the displace-

ment length ratio or displacement length coefficient. It is the

displacement in tons of a vessel similar to the one under con-

sideration and i GO feet long.
2. Skin Resistance Determination. necessary to con- It is

sider separately the two elements of resistance, Skin Resistance

and Residuary Resistance.
The former is the greater in most practical cases and its inde-

pendent calculation is very simple. We have seen that the for-

mula for Skin Resistance is Rf = fSV 1 83

where / is coefficient of
friction from Tideman or Froude, S is wetted surface and V is

speed in knots. For a complete design 5 may be accurately cal-

culated. For a preliminary design it may be closely estimated
from the formula S = c \/DL, where c is the wetted surface
coefficient and may be taken from Fig. 41.
If we were concerned with Skin Resistance only, it would prob-
ably be the best plan always to determine E.H.P./ by formula as
was done when calculating the E.H.P./ of a full-sized ship from the
results of model experiments. But
necessary to use a more it is

complicated system of variables in order to handle Residuary

Resistance, so it is desirable to express Rf in the same variables.
seen that Rf = /SF1 83 and S = c \/T)L. Hence = R

Wejiave f
fc \/DLF
Write y

L, Then y =
o-u 1000000 -- D or D=

Also write x = -^=- Then V = x \/Z F 1 ' 83 =

83 0-915

D -/
Whence finally


In the above / varies slightly with length, L' 085 varies slowly
with length, and c is an almost constant coefficient.
Evidently then for a given length and value of c we can plot

contours of on and .
r ., as primary variables. Fig. 78
shows such contours for a length of 500 feet, the value of /
being taken from Table VI of Tideman's constants. But y^r does

not vary very rapidly with length and it varies with length only.

So Fig. 78 can be applied to all lengths and values of c by the

use of simple correction factors. The correction factors for length
are given on the scale beside the figure to the right. In Fig. 78
the standard value assumed for c is 15.4. If we are dealing with
a vessel for which we know that c is 16.0 for instance, it is obvious
* from
that we should multiply the values of Fig. 78 by-
D 15.4
3. Residuary Resistance from Standard Series. Take up now
the question of Residuary Resistance. Here we are driven to the
use of model results.
Fig. 79 shows the lines used for a series of models which may be
called the Standard Series.

Fig. 79 shows a model having a longitudinal coefficient of .5554,

a midship section coefficient of .926 and a displacement length
ratio of 106.95. The stem was plumb and the forefoot carried

right forward in a bulbous form. From these parent lines a num-

ber of models were constructed with various values of beam

draught ratio, etc.

There were two values of beam draught ratio used, namely
2.25and 3.75.
There were five values of displacement length ratio used, namely
26.60, 53.20, 79.81, 133.02 and 199.52.
There were eight values of longitudinal coefficient used, namely
.48, .52, .56, .60, .64, .68, .74 and .80.

Fig. 80 shows relative curves of sectional area used for the

ight values of the longitudinal coefficient.
Each of the 80 models was run, its curve of residuary resistance
in pounds per ton determined and from the results of the two

groups of different beam ratios after cross fairing, Figs. 81 to 120

were plotted.
Each figure refers to a fixed value of and of =. It shows
" vL
contours of residuary resistance in pounds per ton over the range of

values of longitudinal coefficient and - most likely to be found


in practice. In applying the results of Figs. 81 to 120 for approxi-

mate estimates of E.H.P. for beam draught ratios other than

2.25 and 3.75, interpolation of resistance is linear. This is war-
ranted by results of experiments with models from the same
parent model and of intermediate beam draught ratio. While
not quite exact, it seems sufficiently close to the truth for practi-
cal purposes.

4. Estimates of E.H.P. from Standard Series. We are now

prepared to calculate curves of E.H.P. for a vessel of any size
beam ratio and length within the range covered by Figs. 81 to
1 20 and from the parent lines of the Standard Series. Table X
shows the complete calculations for a vessel of the size, beam

ratio and length of the U. S. S. Yorktown. For each value of =
the corresponding figures for -the two beam ratios are consulted

and columns 2 and 3 filled with the values of

for longitudinal

coefficient =.592 and = 138.1. Then in succession columns


5, 4 and 8 are filled as indicated in the headings. Column 6 is

filled from Fig. 78.


The correction factor (&) for -=* is obtained as clearly indicated

in the heading and column 7 is column 6X6.

The total residuary resistance in pounds per ton is entered in
column 9, and column 10 contains the E.H.P. factor by which this
must be multiplied to determine at once the E.H.P.
This E.H.P. factor is .00307 DV, but it is convenient to call it

.00307 D VL X - = Then (a) or .00307 D Vl, is calculated and

entered in the heading and the values of -=. are found in the
first column. Column n contains the E.H.P. and column 12 the

corresponding values of V. Column 10 could be obtained by


multiplying column 12 by .00307 D, but the methods indicated in

the table will usually be found more convenient in practice.
5. Comparison of Standard Series Estimates with Yorktown
Model Results. As illustrating the application of the Standard
Series results to estimates of E.H.P. attention
invited to Fig. 121.

This shows the E.H.P. curve of the Yorktown as determined by

experiment with a model of the vessel and the curve of E.H.P. from
the Standard Series as calculated in Table X. It is seen that the
Standard Series E.H.P. is less than the actual model E.H.P. up to
the speed of 18 knots, which is higher than the trial speed of the
Yorktown. This simply shows that the Standard Series lines are
better than those of the Yorktown. As a matter of fact, hardly
any models of actual ships tried in the Model Basin have shown
themselves appreciably superior as regards resistance to the Stand-
ard Series and very few have been equal to it. Figs. 76 and 122
show further comparison between actual models and Standard
Series results. Fig. 76 shows six E.H.P. curves calculated from
six actual models for the Yorktown and five variants having the
same length and displacement and derived from the Yorktown
lines but having varying proportions of beam and draught as indi-
cated in the table with Fig. 76.
Fig. 122 shows E.H.P. curves for the same six vessels estimated
from the Standard Series results. It is seen that the agreement is
reasonably close. The Standard Series generally shows less power
than the vessels on Yorktown lines, and the curves from it are
more closely bunched, but the general features of the two figures
are markedly similar.
Effect of Longitudinal Coefficient.
6. Figures 81 to 120, show-
ing the residuary resistance for vessels on the lines of the Standard

Series, are worthy of the most careful and attentive study. Atten-
tion may be called to one or two of the most obvious features.
It is seen that for nearly every speed there is for a given displace-
ment length ratio a distinct minimum of resistance correspond-

ing to a definite longitudinal coefficient. For low and moderate

y =
speeds up to i.i the best longitudinal coefficient is between

.5 and .55. Above this point, however, the optimum longitudi-


nal coefficient rapidly increases, reaching about .65 when - =

y =
and being a little greater still when = 2.00.
The influence of variation of longitudinal coefficient is greatest
below extreme speeds, and it is very great indeed at some speeds.

Thus, in Fig. 91, for = 2.25,

= i.i,
= 100, the resid-
tl VL
uary resistance in pounds per ton for a longitudinal coefficient of .55
is about 6j. But for a longitudinal coefficient of .65 the residuary
resistance in pounds per ton is more than doubled being over 14.
7. Effect of Displacement Length Ratio. The change in type
of the figures with increasing speed length ratio is notable. Thus,
for speed length ratio of .75 the contours are nearly vertical in
wake of the rather full coefficients which such slow ships would
usually have. This means that if we keep length and speed con-
stant and increase displacement, the residuary resistance per ton
remains practically constant or the residuary resistance varies as
the displacement. Consider now
where the speed length
Fig. 100,
ratio is 2.0. For displacement length ratio = 30 the optimum lon-

gitudinal coefficient is about 63 and the residuary resistance in

pounds per ton about 51. For the same longitudinal coefficient
and a displacement length ratio of 50 the residuary resistance in
pounds per ton is about 77. This 77 applies not only to the 20
increase above 30 but to the original 30 as well as that. Though
the relative displacements are as 50 to 30, the relative residuary
resistances are as 50 X 77 to 30 51 or as 3850 to 1530. X So an
increase of displacement of 66 per cent means an increase in
residuary resistance of about 165 per cent.
8. Optimum Midship Section Area. The displacement, length
and longitudinal coefficient being fixed, the area of midship sec-
tion can be calculated without difficulty. For convenient refer-

ence, however, Fig. 123, derived from a series of 2.92 beam

draught ratio on the lines of the Standard Series, gives contours of
/- L \

(midship section area) -5-

( '-) for minimum residuary resistance

plotted on speed length ratio and displacement length ratio. From

this diagram there may be readily determined in a given case
the optimum midship section area as regards residuary resistance.
Of course, in practice there are many considerations affecting
midship section area besides that of minimum residuary resist-
ance, and the midship section cannot be fixed from considerations
of resistance only.

9. Effect of Length. Figs. 81 to 120 do not show directly the

effect of variation of length may be readily utilized to do this.
Thus, suppose it is required to design a vessel of 30,000 tons
displacement to be driven at 29 knots. For preliminary work
B =
assume 3.75.

Assuming various lengths we use Fig. 78 to determine the

corresponding values of the frictional E.H.P. and the Standard
Series figures for = 3.75 to determine the residuary E.H.P.
It is assumed in this preliminary work that it is possible to adopt
the optimum cylindrical coefficients.
Fig. 124 shows
for the case under consideration separate curves
of frictionaland residuary E.H.P. and a curve of their sum, or the
total E.H.P. all plotted on L. The slow growth of frictional
E.H.P. and the rapid falling off of residuary E.H.P. with length
are evident. It is seen that the minimum total E.H.P. corre-

sponds to a length of 950 feet. It has already been pointed out

that in practice the length should be made less than that for mini-
mum resistance.

Thus, if the vessel were made 850 feet long the increase of E.H.P.
would be infinitesimal, and if made 750 feet the increase would be
only from 36,500 to 40,200. As the length is made shorter, however,
the E.H.P. begins to rise very rapidly. This figure illustrates
clearly the enormous effect of length upon residuary resistance.
Thus the residuary E.H.P. is a little over 5000 for a length of
950 feet and is 50,000 for a length a little below 600 feet.
It may be noted here that for a case such as that shown in Fig.

124 it would usually be advisable to adopt a longitudinal coeffi-

cient above that for minimum resistance. This for several reasons,

among which may be mentioned the better behavior in a sea way

associated with the fuller ends, and the better maintenance of

speed in rough water associated with the smaller midship section.

For a vessel where is large, however, it is usually advisable to
make the longitudinal coefficient less than that for minimum re-
sistance. Such vessels are nearly all torpedo boats or destroyers,
which cruise usually at speeds below their maximum, and it is

advisable to save power at cruising speeds by using a longitudi-

nal coefficient a little below that best for maximum speed.
10. ParallelMiddle Body Results. The Standard Series re-
sults of Figs. 8 1 to 1 20 do not apply to one important type of

vessel, namely, the slow vessel of speed length coefficient from .5

to .8 with a parallel middle body. Two questions arise in this
connection. whether as regards resistance it is advisable to
use a parallel middle body, and second, what is the most desirable
length for the parallel middle body in a given case ?
Experiments were made with models having a midship section
coefficient of .96, a ratio of beam to draught of 2.5, various values
of displacement length coefficient and three values of longitudinal
coefficient, namely, .68, .74 and
For each longitudinal coeffi-

cient and displacement length coefficient one model was made

without parallel middle body and four with parallel middle body.
The lengths of parallel middle body expressed as fractions of whole

length were as follows:

For .68 longitudinal coefficient, .09, .18, .27, .36.

For .74 longitudinal coefficient, .12, .24, .36, .48.
For .80 longitudinal coefficient, .15, .30, .45, .60.

Curves of residuary resistance were deduced somewhat as in Figs.

81 to 120.
It was found that at low speeds there is a distinct advantage in

using parallel middle body. This means, of course, that at these

speeds for a given longitudinal coefficient advisable to place
it is

as much displacement as possible amidships and to fine the ends.

It was found too that when contours of residuary resistance
were plotted for a given longitudinal coefficient and speed length

being percentages of parallel middle body

coefficient, the abscissae
and the ordinates displacement length coefficients, the contours were
practically vertical in the vicinity of the optimum length of paral-
lel middle body or that for minimum residuary resistance. In other
words, under these conditions the residuary resistance in pounds
per ton does not vary much with displacement length coefficient
and the latter can be practically eliminated as a variable. Hence,
for the purpose in hand the results of the experiments with the
models of parallel middle body may be summarized in Figs. 125,
126 and 127 which apply to the three cylindrical coefficients used,
namely, .68, .74 and .80. Thus, consider Fig. 126. The abscissae
are values of One curve shows percentage length of parallel

middle body for minimum residuary resistance. The correspond-

ing residuary resistance is given. For convenience, two other
curves are given, which show approximately the percentages of
parallel middle body greater and less than the optimum, which
correspond to residuary resistance ten per cent greater than the
minimum. These give an idea of the variations of length of par-
allel middle body permissible without great increase of residuary
That the saving by the use of parallel middle body is real is
evident from Fig. 128. This gives the three curves of residuary
resistance in pounds per ton for the optimum length of parallel
middle body from Figs. 125, 126 and 127 and average curves for
the same longitudinal coefficients for the Standard Series with no

parallel middle body. The lines of the Standard Series appear to

be slightly superior to those used for the models with middle body,
but even so the saving by the use of the optimum length of par-
allelmiddle body is appreciable.
While three coefficients are not enough to fair in exact cross
curves on longitudinal coefficient, an approximation can be made
from them of ample accuracy for practical purposes, and Fig. 129
shows plotted on speed length coefficient and longitudinal coeffi-
cient by full lines contours of optimum length of parallel middle

body and by dotted lines corresponding residuary resistance in

pounds per ton. It should be understood that the optimum

length of parallel middle body shown in Fig. 129 can be materially

departed from, as indicated in Figs. 125, 126 and 127, without
much increase of residuary resistance.
Particular attention is invited to Fig. 129 which shows how

rapidly residuary resistance increases with speed for full models

and also how rapidly at speeds above the very lowest it increases
with increase of longitudinal coefficient. A judicious selection of
a longitudinal coefficient suitable for the speed is just as impor-
tant for slow vessels as for fast. While hard and fast rules cannot
be laid down, experience appears to indicate that few good de-
signers adopt coefficients and proportions for slow ships such that
the residuary resistance is much over 30 per cent of the total; and
though it is as low as 20 per cent of the total in but few cases,
this figure, if it can be attained for low-speed ships, results in
vessels which are very economical in service.

15. Squat and Change of Trim

In discussing the disturbance caused in the water by a ship,

this question has been touched on, Figs. 45 to 49 showing changes
of trim and level for two models at several speeds.
i. Changes of Level of Bow and Stern. It is the practice at
the United States Model Basin when towing models for resistance
to measure the rise or fall of bow and stern and then plot curves
showing the relation between speed and change of level of bow and
of stern. These results apply linearly to model and ship at corre-
sponding speeds; that is to say, if the ship dimensions are / times
those of the model, the rise of bow of the ship at a given speed will
be times the rise of the model at corresponding speeds.

This fact
is taken advantage of in plotting the curves of Figs. 130

to 139, which show for 10 models curves of change of level of bow

and stern, the departures of bow and stern from original level being

expressed as fractions of length L and plotted not on actual speeds

but on values of -=. These curves are then applicable to any size

of ship upon the lines of the model from which they were deduced.
Actual values of rise and fallcan be determined promptly for any

speed and length of ship by multiplying by L the values of the

-V values of the ship. of trim in

curve ordinates for the Change
degrees can be determined with sufficient approximation by multi-
plying the difference between the scale values of bow and stern
levels by the constant 57.3, the value in degrees of a radian or

unity in circular measure. There are given on the face of each

figure the values of the displacement length coefficient, the longi-
tudinal coefficient and the midship section coefficient of the corre-

sponding model, thus enabling adequate ideas of its general type

to be formed.
The curves of Figs. 130 to 139 show what would happen to vessels
that are towed. The
propeller suction in the case of screw steamers
would cause such vessels when self-propelled to sink more by the
stern than indicated, but the difference would not be great.
2. General Conclusion as to Level and Trim Changes with Speed.
- The results of Figs. 130 to 139 are typical of results shown by
hundreds of other models which warrant the general conclusions
below upon the subject of the change of level and trim of vessels
under way in deep smooth water.
V =
i . At low and moderate speeds below i .o both bow and
stern settle. For short full vessels this bodily settlement is much
greater than for long fine vessels.

V =
2. Below i.o about, there is little or no change of trim.
In the majority of cases the bow settles a little faster than the stern,
particularly for rather full vessels.
V =
3. As speed is increased beyond = i.o the bow settles more
*^ Li

y =
slowly, reaches an extreme settlement at about - 1.15, and

soon begins to rise rapidly, its original level when - y= =

1.3 to 1.4, and continuing to rise. The stern settles more and more
rapidly beyond about = =1.2, and settles much more rapidly

than the bow rises, so that the ship as a whole continues to settle
while rapidly changing trim.
V =
4. At about -=. 1.7 to 1.8 the stern is settling less rapidly than

the bow is rising, so that bodily settlement reaches its maximum.

The stern does not change its level much beyond = = 2.0, while the
bow rises always with increase of speed, the result being that the
y =
vessel is rising again at speeds beyond 2.0 about. The
center of ordinary vessels will never rise to its original level at any

practicable speed; but, since the effect of the passage of the vessel
is to depress theimmediately surrounding water, it may seem at
as if the vessel had risen above its
very high speeds original level.
Vessels of special forms and skimming vessels if driven to extreme

speeds may rise bodny.

3. Critical or Squatting Speed. The most striking feature of

change of level curves is the abrupt change at about

- = = 1.2,

the critical speed at which the bow begins to rise and the stern to
settle abruptly, causing rapid change of trim.
This "squatting" is often thought to be a cause of excessive resist-

ance. As a matter of fact, it is simply a result of large bow wave

resistance. At- - = i.i to 1.2 the first hollow of the bow wave
issomewhere near amidships and the second crest somewhere for-
ward of the stern holding it up, as it were. With increase of speed
the crest moves aft clear of the stern and the hollow moves aft
toward the stern. The stern, of course, drops into this bow wave
hollow, causing the "squatting" or rapid change of trim noticed.
As speed is increased the hollow in turn moves beyond the stern
and the vessel advances on the back of its own bow wave, as it were.
The higher the speed, the longer the bow wave and the closer the
vessel is to the crest.

marked squatting generally means great

It is perfectly true that

resistance, because it is the result of an excessive bow wave with a

deep first hollow. With no bow wave there would be no squatting,


and with slender models having small bow waves squatting is much
less marked than for short full models. In every case, however,
it is a symptom rather than a cause of resistance.
4. Perturbation below Critical Speed. Figs. 131, 132, 133 and
139 show perturbation in the change of level curves below the

critical speed
=r = 1.2. These models are very full ended and

have such strong bow waves that as the hollow corresponding to

V =
1.0 passes the stern it drops into it and the bow rises.
V J^j

Reverse operations take place as the next bow wave crest passes,
and then we reach the critical speed, when the stern drops into the
bow wave hollow corresponding to -=. = 1.2 and over.
Instead of the pronounced perturbations of quite full models
we find for moderately full models the wave hollows and crests

passing the stern at speeds below the critical speed cause the curves
of change of level to have flat or unfair places. Fig. 135 is a case
in point.
For fine models the bow wave is generally so small and the change
of level also so small that no effect of the bow wave can be traced
V =
in the curves until we reach the critical speed = 1.2.

In considering Figs. 130 to 139 we should bear in mind that the

large variations of level and trim shown are for speeds reached by
very few vessels.
The curves of Figs. 130 show changes of level with reference
to 139
to the natural undisturbed water level, and not with reference to the
level of the water in the immediate vicinity of the ship. We have
already seen in discussing the disturbance of the water by a ship
that, as illustrated in Figs. 45 to 49, the passage of the ship causes
disturbances of water level in its vicinity the net result being that
on the average there is depression of the water immediately sur-
rounding the vessel.
The changes of level, trim, etc., shown by under way in
shallow water differ somewhat from those found in deep water, and
will be taken up when considering other shallow-water phenomena.

16. Shallow-Water Effects

1. Changes in Nature of Motion from that in Deep Water. -
It is to be expected that as the water shoals the resistance of a
ship moving through it will become greater. When the water can
move freely past the ship in three dimensions the pressures set up
by the ship's motion would naturally be less than when shallowness

compels the water to motions approaching the two-dimensional

character. Referring to Fig. 21, the greater stream pressures for
plane or two-dimensional motion are evident. In shallow water
these extra pressures cause waves larger than those in deep water,
and in shallow water the lengths of waves accompanying a ship
at a given speed are greater than for the same speed in deep water.
These are the principal factors differentiating shallow-water resist-
ance from deep-water resistance. There is a third factor, namely,
the change in stream velocities past the surface of the ship when in
shallow water. This factor would increase resistance somewhat,
but its effect would seem to be so small that it is not necessary to
consider it since we cannot at present determine with much accuracy

the effect of the dominant factor, namely, the change in wave

production. We can, however, as a result of experiments with
models and full-sized boats get an excellent qualitative idea of the
2. Results of Experiments in Varying Depths. Figs. 140 to
144 show a series of curves of resistance or indicated horse-power.

The data from which these curves were constructed came from
widely separated sources. The information regarding the German
torpedo boat destroyer came originally from a paper by Naval
Constructor Paulus in the Zeitschrift der Vereines Deutsche

Ingenieure of December 10, 1904. Data for the Danish torpedo

boats was given by Captain A. Rasmussen, one of the first experi-
" "
mental investigators in Makrelen
this field. The
data was
given in Engineering of September 7, 1894, and the "Sobjornen"
data in a paper read before the Institution of Naval Architects in
1899. Data for the torpedo boat model was given by Major
Giuseppe Rota, R. I. N., in a paper read in 1900 before the Insti-
tution of Naval Architects, the experiments with the model having

been made in the Experimental Model Basin at Spezia, Italy.

Information from which the curves for the Yarrow destroyer were
deduced was given in a paper before the Institution of Naval
Architects in 1905 by Harold Yarrow, Esq. In Mr. Yarrow's
paper curves of E.H.P. were given as deduced from model
experiments in the North German Lloyd experimental basin at
Each curve refers to a definite depth of water, which has been
expressed as a fraction of the length of the vessel. Furthermore,
speed has been denoted not absolutely but by values of
Deductions from Experimental Results.
3. Examining the
curves, which range from those for a 145-pound model to those for
a 6oo-ton destroyer, and bearing in mind the varying depths ex-

pressed as fractions of the length, we seem warranted in concluding

that in a depth which is a given fraction of the length the perturba-

tions occur at substantially the same values of regardless of

the absolute size. The reason for this must be sought in the rela-
tion between the length of a wave traveling at a given speed in a
given depth of water and length of vessel.
By the trochoidal theory the formula giving wave speed in shallow
water is

*f _
4*^ 27T
6 *+I
where / is length of wave in feet, d is depth of water in feet and v

is speed of wave in feet per second.

Now let L denote length of ship in feet and put I = cL.
Also let V denote common speed of ship and wave in knots.

Then V= v ~; Substituting, reducing and putting g = 32.1 6

we have

Fig. 145 shows contour curves of equal values of c plotted on axes of

7 and -=' Fig. 145 also shows in dotted lines curves deduced
somewhat arbitrarily from Figs. 140 to 144 and other data showing
the loci of the points at which increase of resistance due to shoal
water becomes noticeable, attains its maximum and dies away.
The datais not thoroughly concordant, and the dotted curves of

Fig. 145 should be regarded as a tentative attempt to locate regions,

rather than points. The broad phenomena, however, are clear.
A high-speed vessel in water of depth less than her length will at a
given speed in a given depth begin to experience appreciably in-
creased resistance as compared with its resistance in deep water.
The increase of resistance above the normal becomes greater and
greater as speed increases until it reaches a maximum. This maxi-
mum appears to be at about a speed such that a trochoidal wave
traveling at this speed in water of the same depth is about ii times
as long as the vessel. As the vessel is pushed to a higher speed the
resistance begins to approach thenormal again, reaches and crosses
the normal at about the speed indicated in Fig. 145, and for

higher speeds the resistance in shallow water is less than in deep

was at one time supposed that the speed for maximum increase
in resistance was that of the wave of translation. This, however,
as illustrated in Fig. 145, holds only for water whose depth is less
than .2 L. For greater depths the speed of the wave of translation
rapidly becomes greater than the speed of maximum increase of
There are obvious advantages in the model-basin method of

investigating this subject. Consider, for instance, Fig. 144 showing

actual falling off of resistance beyond the critical speed in the
curves for the Yarrow destroyer which were obtained by model-
basin experiment. This remarkable feature would never be detected
on a full-scale trial of an actual destroyer, because if such a vessel
were forced to surmount the hump it would leap the gap, as it
were, and show a sudden jump in speed. Theoretically if the depth
of water were absolutely uniform it would be possible after the

jump in speed to gradually throttle down until the boat would be


working in the hollow, but the chance of this ever being done, unless
itwere known that the hollow should be there, is infinitesimal.
4. Shallow-Water Experiments at United States Model Basin.
That the hollow really exists, as shown in the curves for the Yarrow
destroyer, confirmed by published results of other model-basin

shallow-water experiments and by a number of carefully made ex-

periments in the United States Model Basin.
Fig. 147 shows curves of resistance and change of trim of the
model of a fast scout in various depths of water. The model was
20 feet long on L.W.L., with 2'. 268 beam and o .842 mean draught.
It displaced in fresh water 996 pounds. The corresponding speed
of the model for 30 knots speed of the full-sized ship would be

only 6. 6 1 knots, but the experiments were carried to a much higher

speed as a matter of interest.
The sudden andpeculiar drops in the shallow-water curves are
very marked. It is seen that they are accompanied by peculiar
corresponding perturbations in the curves showing change of trim
or change of level of bow and stern. We have from Fig. 147 :

Depth of water

spends to the wave of translation, which advances with less de-

mand upon the model for energy to maintain it than was the case
at a slightly lower speed when the wave system was being built
up even ahead of the model.
At the higher speeds the waves are forced waves, necessarily
departing widely from trochoidal waves. It should be remarked
that the high "deep water" resistance of the model at speeds in
the vicinity of 8 knots may be in part due to the limited depth
(14 feet) of the basin, but is probably mostly due to the appear-
ance of the last normal deep-water hump of resistance curves.
The hump which appears below 6 knots in 46 inches depth is

found at about 8 knots in 14 feet depth.

5. Shallow-Water Resistance for Moderate and Slow Speed

Vessels. The case of greatest practical interest is that of the

vessel of moderate speed say capable of a deep-water speed in
knots of .9 \/L or less. Such a vessel in shallow water cannot
be pushed beyond the last hump of her resistance curve,and hence
always loses speed in shallow water. For such vessels we would
like to know the least depth of water in which resistance is not
appreciably increased or speed appreciably retarded and the
amount of increase of resistance in water that is shallower.
Results of experiments bearing directly on the first question
were published in 1900 in a paper before the Institution of Naval
Architects by Major Giuseppe Rota. Major Rota experimented
with models of five vessels, one being the torpedo boat model,
whose results are given in Fig. 143. Each model was run in vari-
ous depths of water and the results carefully analyzed for the pur-
pose of determining the depth at which increased resistance began.
For the purpose of analysis and deducing results applicable to
other vessels important to determine in connection with such
it is

experimental results the fundamental variables, as it were. For

instance, in this case shall we connect the depth of water with the
length, the beam or the draught of the ship? We
have seen that for
high-powered vessels we were led to the use of the ratio between
depth water and length of vessel, which gives satisfactory re-
sults as regards determination of critical points, etc. Considera-
tion, however, appears to indicate that for the vessel of moderate

speed it would probably be better to use the ratio between depth

of water and mean draught of ship, allowing the length factor to
come in through the speed-length coefficient.
While Rota's models could, of course, each be expanded to rep-
resent any number of ships, he gives one size of ship for each as
shown in the table below.

Model No .

The formula, however, has been found to apply satisfactorily to

models of block coefficient higher than .5 tested in the United
States Model Basin. One model of block coefficient slightly above
.65 was tried in various depths and the formula found to apply

To sum up, I think that the above formula from Rota's experi-
ments may be confidently applied :

1. To vessels not of abnormal form or proportions up to a

block coefficient of .65.
2. For speeds for which ~=is not greater than .9.

The formula may be of use beyond the limits indicated above,

but in such cases needs to be applied with caution and discretion.
6. Trial Course Depths. As illustrative of the little impor-
tance attached to this question until a comparatively recent date,
Major Rota paper states: "Stokes Bay, where British
in his 1900

ships used to undergo their speed trials, is only 59 feet deep; the
official measured mile at the Gulf of Spezia, Italy, is about 62 feet

deep; the measured miles at Cherbourg and Brest are 49 and 59

feet respectively." Such depths are now regarded as entirely
inadequate and no speed trials of large ships are regarded as
accurate unless made in deep water. Curiously enough, however,
as indicated in Fig. 145, the shallow course exaggerates the speed
of the very fast vessel, and there are many torpedo craft in exist-
ence whose full-speed trials were held on shallow courses with
resulting speeds greater than would have been attained in deep
7. Percentage Variations of Resistance in Shallow Water.
Coming now to the question of the actual increase of resistance
of a given vessel in water of a given depth, it is necessary again
to make a distinction between the vessel of very high power and

speed and the vessel of moderate speed. For the former it is

probably best, as before, to use as the governing variable the ratio
between depth and length, For the latter it still seems best

to use the ratio between depth and draught,^- For either type,

expressing the speed by > we are able for each vessel or model

for which there is adequate experimental information to draw con-

tours on = and ratio between depth and length or depth and
draught as the case may be, which show percentages of increase
over deep-water results. For the very high-speed vessels percent-
ages of decrease will also appear. This work at best can be only
a tolerably good approximation, and hence we assume in it that
the law of comparison applies fully to the total model resistance.

Figs. 148 to 153 are percentage increase diagrams, the type of

vessel being indicated in each case.
The diagrams for the high-speed vessels show percentages of
decrease. For the moderate-speed vessels the percentage increase
of resistance goes up rapidly with
increase of displacement length
coefficient. While Figs. 151, 152 and 153 cannot be said to cover
the ground as would be desirable, they will be better than nothing
and of help in many cases.
Inland navigation ismostly smooth-water, shallow-water navi-
gation, and there is great need of a complete investigation into
the features of form affecting shallow-water resistance. While we
know quite well the general features of the form best adapted to
speed in deep water in a given case we do not know the same thing
for shallow water. It appears probable, however, that if we
wish to make 12 knots in shallow water and are considering vari-
ous models, that one which will drive easiest in deep water at a
higher speed say 1 5 knots or so will drive easiest in shallow

water at the i2-knot speed. If high speed is to be attempted in

inland navigation there are practical advantages in length which
would be excessive for deep-water work. Wave making, with the
resulting wash at banks and piers, should be kept as low as possi-
ble for boats in river service.
8. Shallow- Water Influence upon Trim and Settlement. Fig.
147 shows the curves of the settlement of bow and stern of a scout
model in shallow water. It is seen that the shallower the water
the lower the speed at which marked change of trim begins, and
within the limits of practicable speed the greater the change of trim.

For speeds above those at all possible the trim changes would not
very greatly depart from those for deep water. We are more con-
cerned in practice, however, with settlement and change of trim
at low speeds, corresponding to those at which shallow channels
would be traversed. Fig. 147 shows that at such speeds the effect
of shoal water is simply to increase the settlement of both bow
and stern. In its broad features, Fig. 147 is fairly typical of
change of trim results in shoal water for a number of other models.
We may say that the effect of shoal water upon a vessel under
way is to increase the natural settlement of both bow and stern
at low speed. The
shallower the water the lower the critical speed
at which squatting or excessive change of trim begins and the

greater the change of trim. At high speeds the shallower the

water the more the stern settles and the more the bow rises. At
extreme speeds, however, the stern does not appear to settle or
the bow to rise so far as in deep water. It is interesting to note
in Fig. 147 the peculiar perturbations in the change of level curves
and the evident close connection between them and the remark-
able drops in the resistance curves.

9. Increase of Draught in Shallow Channels. In practice

there are very few vessels of sufficient power to attain high speed
in shallow water, and those that have the power would very sel-
dom use it in shallow water, so that the behavior of vessels as

regards settlement under way at moderate speed in shallow chan-

nels is of more practical importance than their possible behavior
at excessive speeds.
A very interesting investigation of this question was made in
connection with the channel of New York Harbor, and was de-
scribed in detailby Mr. Henry N. Babcock in Engineering News
for August 4, This channel was constantly used by large
steamers passing in and out with very little to spare between their
keels and the bottom of the channel. There were repeated com-
plaints from such vessels that they had touched bottom in places
where the officers in charge of the channels were unable to dis-
cover spots shoaler than the still-water draught of the steamers.
The observations were confined to large transatlantic steamships
passing out of New York, averaging over 550 feet in length. They

were made at three points, one where the channel was 80 to 100
feet deep, one where the low- water channel depth was from 31.1 to

32.5 feet, and a third where the low- water depth was from 31 to
34.5 feet.
The general scheme of most of the observations was to deter-
mine the height above water of marks on the bow and stern before
the steamer left her pier. Then as the steamer passed the observ-

ing station the level of these marks was determined with reference
to the station, and as soon as possible after the passage of the
vessel the water level was determined with reference to the observ-
ing station. Considering all the circumstances, exact observations
are obviously not possible, but after making ample allowance for

possible errors of observation Mr. Babcock's report demonstrates

conclusively that vessels of the type considered when under way
in channels settleboth at bow and stern, and the shoaler the water
and higher the speed the more they settle. It was not practicable
from the results to formulate fully conclusions connecting amount
of settlement with size and type of vessel, speed and depth of
water, but Mr. Babcock, upon analyzing the results, concluded that
for vessels of the large transatlantic steamship type the increase
of draught in feet, when still water clearance under their keels was
less than about 10 per cent of the draught, would be i the speed
of the ship in miles per hour. For a natural clearance of some 30
per cent of the draught the increase in feet would be about iV the
speed of the ship in miles per hour, and for intermediate clearances
intermediate fractions should be used.
Further observations of the character reported by Mr. Babcock
on the settlement of vessels under way, not only in shallow channels
but in canals, would be of much interest and practical value.

17. Rough- Water Effects

i. Causes of Speed Reduction. The effect of rough water upon

speed is like the effect of foulness of bottom almost impossible to
reduce to quantitative rules. The very real and material reduc-
tion of speed of vessels in rough weather is of universal experience.

This, however, is not always due to increased resistance alone.

The motion of the ship may render it impossible to develop full

power. The danger of racing may render it inadvisable to use

full power. The disturbance of the water reduces the efficiency of
the propellers. The conditions may render it impossible to use
full speed without risk of dangerous seas coming on board.
2. Features Minimizing Speed Reduction. The increase of
resistance in rough water is under practical conditions largely a
question of absolute size. Waves 150 feet long and 10 feet high
would not seriously slow a 4o,ooo-ton vessel 800 feet long.
A vessel of a few hundred tons 120 feet long would find them a
very serious obstacle to speed. Pitching enters into the question
of rough-water speed as a very important factor.
When conditions are such as to produce severe pitching, speed goes
down very rapidly. Pitching exaggerates nearly all causes of speed
loss. Not only is actual resistance rapidly increased but racing is
caused, the propeller loses efficiency and more water comes on board.
If it were possible to devise a vessel which would not pitch it

would lose much speed in rough water than one that does

pitch; but though many naval architects have strong opinions on

the subject there is no agreement among them as to the features
of model which minimize pitching. The preponderance of opinion
is probably in favor of the U-bow type and rather full bow water
lines. But pitching is unfortunately largely a question of condi-
tions. Under certain conditions of sea, course, and speed one type
may be superior and under slightly changed conditions distinctly

Apart from absolute size there appears, however, to be one

broad consideration which is of some value as a guide. Suppose
we have two 2o-knot vessels, A and B, of about the same power
and such that at 22 knots A offers distinctly less resistance than B.
There is little doubt that on the average A would lose less speed
in rough water than B.
When for a vessel intended
for a certain service it is necessary to

allow in the design for the effect of rough water upon speed there
is only one safe method to follow namely, to allow a reduction
from smooth-water trial conditions to rough-water service condi-
tions based upon actual experience with previous vessels in the

1 8. Appendage Resistance
1. Appendages Fitted. Substantially all that has been said
about resistance hitherto refers to the resistance of the main body
or hull proper. There are found on actual ships appendages of
various kinds, such as rudders, bar keels, bilge keels, docking keels,
shaft swells, shafts,shaft struts, propeller hubs and spectacle

frames, or shaft brackets or bosses. Shaft tubes, or removable

tubes around the outboard shafts, are seldom fitted nowadays.
The appendages fitted vary. Thus, a single-screw merchant
ship with flat keel will have practically no appendage except
the rudder, the slight swell around the shaft having hardly any
effect. For such a vessel the appendage resistance would seldom
be as much as 4 or 5 per cent of the bare hull resistance.
A twin screw vessel with large bilge and docking keels and
perhaps two pairs of struts on each side may have an appendage
resistance as much
as 20 per cent of the bare hull resistance.

Appendage resistance is largely eddy resistance and can be kept

down to the minimum only by very careful attention to details and

the application of adequate fair waters wherever needed.
2. Resistance of Bilge and Docking Keels. Bilge keels and
docking keels should follow lines of flow and be sharpened at
each end. When this is done it is generally found in experiments
upon models that the additional resistance due to them is not
greater than that due to the additional surface alone. In fact
the additional resistance is sometimes found to be less than that
due to the additional wetted surface. Mr. Froude found a similar
result in his full-sized
Greyhound experiments. While if bilge
keels and docking keels are properly located and fashioned the
additional resistance may be taken as that due to their wetted sur-
face only, the wetted surface they add is often very considerable.
In models bilge keels may be located at appreciable angles
with the natural lines of flow without greatly augmenting resist-
ance beyond that due to their surface, but it does not follow that

the same result would be found in the full-sized ships. It is

necessary to be cautious in applying the Law of Comparison to

eddy resistance. There is little doubt that the law applies to the

Eddy Resistance behind a square stern post, for instance. Here

the eddying for model and ship is found in each case over cor-

responding areas.
But in the case of a bilge keel located across the lines of flow
we may readily conceive that there may be but little eddying
around the model bilge keel and a great deal around the full-
sized bilge keel. This because the pressure of the atmosphere re-
maining constant the total pressure around the full-sized bilge
keel not increased in the proportion required to insure com-

pliance with the Law of Comparison.

3. Resistance of Struts. Probably struts and spectacle frames
are the appendages to which the most careful attention must be

paid from the point of view of resistance. Experiments with a

number of strut arms of elliptical section appear to indicate that
the resistance in pounds per foot length may be expressed with
fairapproximation for areas from 40 square inches to 175 square
inches by the following semi-empirical formula:


Where R is resistance in pounds per foot length, V is speed

through the water in knots and A is area of cross section of strut
in square inches. The coefficient C depends upon the ratio be-
tween B, the thickness of the strut section, and L, its width in
direction of motion. The table below gives values of C for vari-

Even not very often reached in practice, the tend-

this ratio is

ency apparently being to make strut arms much narrower and

thicker than they should be.
As regards shape of section, model experiments indicate that a
pear-shaped section, or a section of rounding forward part and
sharp after part, offers the least resistance. Such a section may
show model resistance as much as 10 per cent below the elliptical
There is doubt, however, whether this holds for full-sized struts
for high-speed vessels. Study of Fig. 16 would seem to indicate
that at sufficiently high speeds there must be eddying over all the
rear half of any strut, in which case the thickness of the strut
should be reduced to a minimum. From this point of view, if a
strut of given width and area is to have the minimum thickness
for a given type of head the rear portion should be made of paral-
lel thickness and cut off square. Furthermore, from this point
of view, if air were piped to the rear of a strut the resistance

would be decreased. This question of strut resistance is worthy

of further careful experimental investigation. Pending this, the
approximate formula and coefficients above for elliptical struts

may be used, and it may be assumed that the elliptical form is

about as good as any. For moderate speeds the rear portion of
the strut may be brought to a sharp edge, but for high speeds
this refinement will probably be of little use.

4. Resistances of Propeller Hubs. Behind the strut hub the

propeller hub is fitted, and for propellers with detachable blades
is usually larger than the strut hub. About all that can be done
hub is to fit a
for the propeller conical fair-water behind it. Model
experiments show that a long fair-water, say of length about twice
the diameter of the propeller hub, offers materially less resist-
ance than a short fair-water of length say about one-half the
diameter of the propeller hub.
While there is some doubt whether the long fair-water would
show up so well in comparison on the full-sized ship, the length of
fair-water should not be skimped.
With quick running propellers the objections to large hubs have
become more evident and there is a tendency to use solid pro-

pellers with small hubs. From the point of view of appendage

resistance, these are distinctly preferable to large hubs.

Resistance of Spectacle Frames or Propeller Bossing.
5. In
merchant practice, struts are not much used for side screws,
being replaced by spectacle frames or propeller bossing.
These appendages, if well formed, offer less resistance than thick
struts with the bare shafts, etc., but in many cases wide, reason-
ably thin struts would offer less resistance than shaft bosses.
Shaft bosses are, however, usually regarded as giving better secu-
rity to the shaft, and certainly give access to a greater portion of
its length. They absorb much more weight than struts. The
angle of the web of a shaft boss may vary a good deal from what
may be called the neutral position, or position where it is edge-
wise to the flow over the hull without very great effect upon the
model resistance, but there is a little doubt that the full-sized
ship will be prejudicially affected if the shaft boss webs depart too
far from the neutral position. Eddying is liable to appear in the
case of the full-sized ship which does not occur in the case of the
The angle of such webs has a powerful influence upon the stream
line motion in the vicinity of the stern. A vertical web or a
horizontal web tends seriously to obstruct the natural water flow
and drag more or dead water behind the ship. It seems to

be usually the tendency from structural considerations to work

the shaft boss webs somewhere near the horizontal. From the
point ofview of resistance alone a 45 angle for the rear edge may
not be too great. This is another case where conflicting consider-
ations necessitate a compromise. The determination of after lines
of flow over the hull will greatly facilitate the determination of
the most suitable shaft boss arrangements.
6. Allowance for Appendages in Powering Ships. In esti-
mating from model experiments the effective horse-power of a
ship with appendages the methods are the same as for the bare
hull. From the total model resistance the frictional resistance for
the total wetted surface including appendages is deducted and the
remaining or residuary resistance treated by the Laws of Compari-
son. From what has been said in discussing appendage resist-

ance, it is evident that estimates of E.H.P. with appendages are

apt to be less accurate than estimates of the net or bare hull
E.H.P. unless care has been taken so to shape appendages that
they do not develop in the full-sized ship eddies which have no
corresponding eddies in the case of the model.
In practice, it is customary and almost necessary to power a
new design from model experiments with bare hull only. This is
readily done by using for the ratio between the bare hull E.H.P.
and the I.H.P. of the ship with appendages a conservative coeffi-
cient of propulsion based upon coefficients of propulsion actually
obtained from past experience with vessels reasonably similar as
regards appendages to the case under consideration.


19. Nomenclature Geometry and Delineation of Propellers

i. Definitions and Nomenclature. A screw propeller has two

or more blades attached at their inner portions or roots to a hub
or boss, which in turn is secured upon a shaft driven by the pro-

pelling machinery of the ship. Figs. 154 to 157 show plans of a

three-bladed propeller for a naval vessel. This is a true screw -
that is, the face or driving face is a portion of a helicoidal surface
of uniform pitch. A helicoidal surface of uniform pitch is the
surface generated by a line the generatrix at an angle with
an axis which revolves about the axis at a uniform angular rate
and also advances parallel to the axis at a uniform rate. A
cylindrical surface concentric with the axis will cut such a heli-
coidal surface in a helix. The pitch of the helicoidal surface is
the distance which the generatrix moves parallel to the axis dur-
ing one complete revolution. Figs. 154 to 157 show a three-
bladed right-handed propeller that is, a propeller which, viewed
from aft, revolves with the hands of a watch when driving the
ship ahead. The various portions of a propeller are indicated in
the figures, such as the face and back of the blades, the leading
edge and the following edge, the tip and the root. Since in prac-
tice the back of each blade is its forward surface, care must be
taken to avoid confusion.
This result will be obtained by avoiding such expressions as
" "
forward face," and adhering to the terms
after face," etc.,
"face" and "back." The word "face" will always denote the

driving face or the face which pushes the water astern when the
" "
propeller is in action, while the word back naturally denotes
the surface opposite the face.
While a true screw as already indicated is a screw propeller

whose blade faces are all portions of helicoidal surfaces of the

same pitch, there are many variants from the true screw.
Each point of the face may have its own pitch, which may be
denned as the distance parallel to the shaft axis which an ele-
mentary area around the point would move during one revolu-
tion around the shaft if it were connected to the shaft by a rigid
radius and working in a solid fixed nut. Fig. 158 shows two views
of a small elementary area LL connected to the shaft axis O by a
radius r. This area makes an angle with the perpendicular to the
axis called the pitch angle and denoted by 6 in Fig. 158. If p

denote the pitch of LL, during one revolution in a solid nut its
center would advance along the helix OCCD, to the point D at
a distance p along the axis from 0. If then we unroll the cylinder
of radius r, upon which has been traced the helix OCCD, this
helix will become the straight line OP of Fig. 158, while = p, PM
the pitch.

OM = 2 wr and tan 6 = 1>

2 irr

There are several typical variations of pitch which are used

more or less for actual propellers. Thusthe pitch increases as

we pass from the leading to the following edge, the blade is said to
have axially increasing pitch. If the pitch increases as we go out-
ward, the blade is said to have radially increasing pitch. If the
pitch decreases as we go outward, the blade has radially decreasing
pitch. A blade may have pitch varying both axially and radially.
Pitch of the blade face only has been considered in the above,
and in an ideal blade of no thickness that is all that need be con-
sidered; but for actual blades we need to consider the pitch of the
back of the blade as well. Evidently each point of the back of
an actual blade has a distinctive pitch. For blades such as shown
in Figs. 154 to 157, where the face has uniform pitch and the blade
sections are of the usual ogival type, the pitch of the center of
the blade back is the same as the pitch of the face. The pitch of
the leading portion of the back is less; and of the following por-
tion greater than the face pitch. These pitch variations over the
blade back have important effects upon propeller action.
The ratio between pitch and diameter is called pitch ratio, and

the ratio between diameter and pitch is called diameter ratio.

Each point of a blade has, of course, its own pitch ratio and
diameter but these expressions are also used in reference to
the propeller as a whole. When so used the diameter referred to
is the diameter of the screw or of the tip circle, and the pitch is

the uniform pitch of the face for a true screw and an assumed
average face pitch for a screw of varying pitch.
There are two other ratios which it is convenient to define here.
Fig. 159 shows a radial section through the center of a blade of

very common
type by a plane through the axis. This plane in-
tersects back and face of the blade in two straight lines, which,

prolonged through the hub to the axis, cut it at C and A respec-


CA -
The ratio is called the blade thickness ratio and is
evidently constant for similar propellers, whatever their size.
The blade section in Fig. 159 is shown raking aft, the total rake
reckoned along the mid-thickness of blade sections being in the

figure BO. Then

- is called the rake ratio. It is reckoned

positive for after rake and negative for forward rake.

Propellers do not in practice move through the water as through
a solid nut. They advance a distance less than their pitch for
each revolution. Under given conditions of operation the distance
advanced is the same for each revolution, hence the path of each
element is a helix and can be developed into a straight line. Recur-
ring to Fig. 158,
= OC\C\D\ is the helical path of LL with slip and
OS the development of this helix. As before, POM is the pitch
angle 6. The angle POS is called the slip angle and will be denoted
by <f>. Fig. 158 may also be regarded as a diagram of velocities,
OM being the transverse or rotary velocity of the element and MS
its velocity parallel to the axis. MS often called the speed ofis

advance, and MP, or the speed for no slip, is called the speed of
the propeller, being the pitch multiplied by the revolutions. Then
PS is the speed of slip or the slip velocity. Slip is usually char-
acterized, however, by the ratio --L ,
or the ratio between the speed

of slip and the speed of the propeller. This is properly called the
slip ratio, or slip fraction. It is also commonly and conveniently
called simply the slip and expressed as a percentage instead of a
decimal fraction. Thus when we say, for example, that a propeller
works with a slip of 15 per cent we mean that

Speed of Propeller Speed of Advance _

Speed of Propeller

Sometimes we need the ratio

Speed of Advance

Speed of Propeller
and thisconveniently be designated the speed ratio.
2. Delineation. In practice a propeller is usually delineated
as in Figs. 154 to 157, by projections of the blades in at least two
directions, an expansion of a blade and sections of a blade.
Views and sections are also shown as necessary to determine the
hub of propeller with solid hubs and the hub and blade flanges
and bolting of propellers with detachable blades.
be observed that the faces of the sections in Fig. 155 all
It will
radiate from a fixed point on the axis, called the pitch point. This
is a more or less convenient arrangement. Referring to Fig. 160,
suppose p is the pitch of a blade at the radius OA = r. Lay off
OP = -
Then tan OAP = ~ + r = -- But from Fig. 158
2 TT 2 TT 2 irr

= tan 6 where 6 is the pitch angle or the angle which the

2 irr

element makes with a transverse plane. Hence in Fig. 160 OAP

and the corresponding angles at the other radii are the pitch angles
at the radii in question.

Figs. 154 to 157 refer to an ordinary true screw of oval blade

contour with a rake so small that it is practically negligible.
Much more complicated forms are used sometimes, the complica-
tions involving varying pitch, curved radial sections, extreme rake
forward or aft, lopsided or unsymmetrical blade contours, and
various types of blade sections. Some forms of propellers are
difficult problems in descriptive geometry. There does not seem
to be any benefit in practice from complicated forms of propellers

and no attempt will be made to take up the problems of their


3. Area and its Determination. The question of propeller area

is a very important one. There are various areas considered in
connection with a propeller. When we speak of the blade area of
a propeller we generally mean what is called the helicoidal area, or
the actual area of the helicoidal faces of the blades. As it happens,
however, a helicoidal surface cannot be developed into a plane so
the helicoidal area of a propeller cannot be determined exactly.
The area we determine is what is called the developed area, the
blade face being developed into a plane by a more or Jess approxi-
mate method.
The disc area of a propeller is the area of the circular section of
its disc or the area of the circle touching the blade tips.
The projected area is the area of the projections of the blade
faces upon a transverse plane perpendicular to the axis.
The ratio between the developed and disc areas of a propeller
is sometimes called the disc area ratio.

The ratio Projected Area -j- Disc Area

is also frequently used

and is of more practical value than the ratio Developed Area -4-
Disc Area.
While the helicoidal face of a propeller blade cannot be developed
exactly into one plane it can be so developed with such slight
distortion that the resulting surface is an approximation amply
close for practical purposes.

Suppose we cut the helicoidal surface of a blade face by a cylin-

der concentric with the axis. It will cut a helix from the helicoidal
surface. If now we pass a plane tangent to the helicoidal surface
at its center, it will cut the cylinder in an elliptical arc. If then

we take that portion of this elliptical arc whose rearward projec-

tion is the same as that of the actual helix of the blade face we will
have an arc of very nearly the same length as the helix. Then if

we take a series of such arcs, swing them into a common plane and
join their extremities by a bounding curve, we shall have a devel-
oped surface which is very close to the actual helicoidal surface
in area.

Fig. 1 60 shows the construction, is the center, P the pitch


point, OA the radius of a cylinder. Let BBB be the projected

blade. Then the cylinder of radius OA cuts BB at C. The plane
atA tangent to the helicoidal surface makes with the axis the angle
OP A the complement of the pitch angle. The minor semiaxis
of the ellipse which it cuts from the cylinder is OA. The major

semiaxis is = ~r~ OA 7 = AP. Draw the elliptical arc AD with

sin OP A

major axis of length AP and minor axis OA length and


position. Then draw the horizontal line CD meeting the ellipti-

cal arc at D. D
a point on the developed blade, and by deter-

mining a series of such points and drawing a line through them

we obtain the developed contour EDBEE. Suppose now we draw
AF horizontal through A and make AF equal in length to the
elliptical arcAD. A line through a series of points such as F will
give what may be called the expanded contour. It is denoted in
the figure by HFBHH. The developed area is usually taken as
BEEKEDB. The expanded area, BHHKHFB, is very close to

the developed area.

The developed area obtained by the above method is slightly
smaller than the true area. The elliptical arcs are not very easy
to draw in practice and a simple method is to use arcs of circles
with radii which are the radii of curvature of the ellipses. Thus
draw PM at right angles to AP and cutting AO produced at M.
Then Mis the center of curvature of the ellipse at A and instead ,

of drawing the ellipse we may draw a circular arc of radius MA.

The developed area thus determined is slightly greater than the
exact helicoidal area, the area using the exact ellipses being
slightly less. But the area determined using the circular arcs is
a closer approximation to the true area, particularly for broad
we generally assume the developed contour, making
In practice
it any desired shape, deduce the projected contour by reversing
the method of development described above, and from the pro-

jected contour deduce by the methods of descriptive geometry the

other projections desired. A very common and very good con-
tour for the developed blade is an ellipse touching the axis, having
the radius as major axis and the expanded breath of blade at

mid-radius as minor In the vicinity of the hub the ellipse is


departed from as necessary to make a good connection.

4. Coefficients of Area for Elliptical Blade. Fig. 161 shows an
elliptical developed blade contour with major axis equal to the
propeller radius. The radius of hub is tV that of the blade.
There is shown dotted a rectangular area touching the hub and
tip circle and of width such that its area is the same as that of the
elliptical blade outside the hub. Then the width of this rectangle
is called the mean width of the blade.
It is convenient usually to use the diameter as the primary
variable when dealing with propellers, so we naturally express the
mean width as a fraction of the diameter.
The ratio (mean width of blade) -r- (diameter of propeller) is
called the mean width ratio and is denoted by h.
This mean width ratio characterizes a blade very definitely and
it isconvenient to express many other features by its use. For the
elliptical blade with hub diameter &
of the propeller diameter let
I denote the maximum width or minor axis of the ellipse. Then
we have mean width ratio = h =.842 - or / = 1.188 hd.
If n denote the number of blades we have the total blade area
or Developed Area = .4 n<Ph.
The projected area for a given developed area depends upon the
pitch ratio, which denote by a. For values of a found in practice,
say from a = .6 to a
= 2.0, the projected area for the elliptical-
bladed propeller of hub diameter .2 of the propeller diameter is
given with close approximation by the formula,
Projected Area
= (0.4267 0.09160) nd?h.
From the above we have the following addit onal ratios for
values of a between .6 and 2.0:
Developed Area = 1.067
Projected Area -j-
Developed Area -f- Disc Area = .509 nh.
Projected Area -j- Disc Area = (.543 .11660) nh.
Fig. 162 shows contours of the ratio (Projected Area)
(Disc -5-

Area) for three-bladed propellers.

While the above formulae and Fig. 162 apply strictly only to
propellers with elliptical blades and hub diameter tV of propeller

diameter, they are accurate enough for practical purposes for any
other hub diameter likely to be found in practice and are rea-
sonably good approximations for any blades of oval type.
5. Twisted Blades. Propellers with detachable blades nearly
always have them fitted so that they can be twisted slightly in the
boss, thus increasing or decreasing the pitch. The blade flange
holes are made oval, as shown in Fig. 156. The twist or rotation
of the blade is about a line or axis through the center of the flange
perpendicular to the shaft.
All pitch angles on the changed a uniform amount.
axis are
For points of the blade away from the axis of twist the change
is less, and for points of the helical surface a quarter of a revolu-

tion from the axis, if the surface were so great, there would be no

change of pitch due to twist. For usual width of blade, however,

the change in pitch angle is practically uniform over the blade
and equal to the angle of twist. Hence the change of pitch due
to twist will be investigated on this assumption.
Let y denote the diameter ratio, 6 the pitch angle at a given

point of radius r and pitch p. Let 7 denote the angle of twist

and y the new diameter ratio after twisting.

Then tan 6 = -*- = - y = - cot 0,

2 irr iry TT

tan (e + 7) = -^7,

cotgcot T-
y =-cot(8+y)=-
= I rrycoty- I
= ycot 7 -7r
TT TT COt 6 + COt 7 TT
Try + Cot 7 iry -\- COt 7

From the above formula, given y and 7, we can readily calculate y'.
For a positive twist or value of 7 the new diameter ratio is less
than the old, the new pitch and pitch ratio being greater. For a
negative twist the opposite holds.
The shown graphically in Figs. 163 and 164. In
results are

Fig. 163 the results are plotted upon diameter ratio. For each
value of 7 a curve is drawn showing the new values of diameter
ratio plotted as ordinates over the old values as abscissae. Con-

tours are shown for each degree of positive and negative twist up
to 6.
Fig. 164 gives the same information as Fig. 163, but the results
are plotted upon pitch ratio.
and 164 illustrate the relative advantages and disad-
Figs. 163
vantages of pitch ratio and diameter ratio when used as primary
variables. Fig. 163 using diameter ratio, once the conception
of diameter ratio is firmly grasped mentally, is simpler and
more readily understood. This is largely because the diameter
ratio at the tip of the blade is the natural starting point, and for

any point of less radius the diameter ratio decreases directly as the
radius. The conception of pitch ratio is more readily formed, but
starting with the pitch ratio of the tips the pitch ratio increases
inversely as the radius and becomes infinite for zero radius. In
either case the tip value is a simple quantity of numerical value
ranging in practice from .5 to 2. When using diameter ratio for
any one blade the field covered, neglecting the hub, is that between
zero and the tip value. When using pitch ratio the field is that
between infinity and the tip value.

20. Theories of Propeller Action

i. Principles of Action Common to all Theories. There have

been a great many different theories of propeller action propounded,
but none which has been generally accepted as agreeing fully with
the facts of practical experience.
The principles underlying the chief English theories of propeller
action are comparatively simple. The resulting formulae are more or
lesscomplicated, but not difficult to apply. In any theory in con-
nection with which mathematical methods are to be used it is almost
necessary to regard the blade as having no thickness. Fig. 165,
which partially reproduces Fig. 158, indicates the motion of a small
elementary plane blade area of radius r, breadth dr, in a radial
direction and circumferential length dl. Looking down we see
this element with its center at 0. If w
the angular velocity of

rotation of the shaft, the transverse velocity of the element is

cor. AOB
is the pitch angle 0, BC the slip and BOC the slip angle

<. We know that tan 8 = p '

Considering Fig. 165 as a
2 TTf

diagram of instantaneous velocities, the line OA or cor represents

the transverse velocity of the element. If there were no slip, the
actual velocity would be parallel to OB since BOA = 6. Then
AB would denote the axial velocity.

AB = OA tan 6 = cor tan = cor

2 TTT 2 7T

When the transverse velocity of the element is un-

there is slip

changed, but the axial velocity is the speed of advance AC, which
is denoted by V A BC is the slip and AB, the speed of the screw,

is thesame as the speed of advance when the slip is zero.

Denote the slip ratio by s.

Then s
= BC = AB AC _2^
_ '__<P- 2vVA_i_y
2 7T

Whence the speed of advance VA*= (i s) BC = s ^*-

2 7T 2 7T

If we take w
as angular velocity per second and r in feet, then OA
or the transverse velocity is in feet per second, and hence all other
velocities are in the same units.
Then we have
Velocity of blade element in the direction of the perpendicular to

its plane
= CD = BC cos = s ^- cos 6.
2 7T

Axial or rearward component of above velocity = CE = CD

cos 6 = s ^- cos2 0.
2 7T

Transverse component of above velocity = DE = CD sin 6 =

s -

sin cos 6.
2 7T

2. Three English Theories of Propeller Action. There are

three theories of propeller action whose detailed consideration
will be of value. They are all contained in papers before the
Institution of Naval Architects. The first was by Professor Ran-

kine in 1865, the second by Mr. Wm. Froude in 1878 and the
third by Professor Greenhill in 1888.
Rankine's fundamental assumption was that, as the propeller
advanced with slip BC, all the water in an annular ring of radius
r was given the velocity CD in a direction perpendicular to the
face of the blade at that radius. Then, from the principle of
momentum, the thrust from the elementary annular ring is pro-
portional to the quantity of water acted upon in one second and
to the sternward velocity EC communicated to it.
Froude considers the element as a small plane moving through
the water along a line OC which makes a small angle with OB, <

the direction of the plane. Then Froude takes the normal pressure
upon the elementary area which gives propulsive effect to vary as
the area, as the square of its speed OC, and as the sine of <j>
the slip

Greenhill makes a somewhat artificial assumption. He assumes
that the propeller is
working end tube. The
in a fixed closed
resultis that the motion communicated to the water is wholly

transverse and would be represented by CF in Fig. 165. The

blade is first assumed smooth, so that the pressure produced by

the reaction of the water is normal to the blade and has of course
a fore and aft component which gives thrust. In all three theories
the loss by friction is taken as that due to the friction of the

propelling surface moving edgewise or nearly so through the

3. Relation between Direction of Pressure and Efficiency. -
Neglecting friction for the present it is evident that all three
theories start with a certain normal pressure. It follows that if

thisnormal pressure be resolved into its axial and transverse com-

ponents, say dT and dQ, we have

41 ~= n =~ OA ~
= 2L =
= 2jrr
dQ AB co
2 TT

Hence pdT = 2 -n-rdQ.

Now 2 TrrdQ = totalwork done during one revolution and hence,

neglecting friction, pdT = total work done during one revolu-

tion. Now the useful work = dTp (i s), hence the efficiency

It follows that, neglecting friction, if the reaction
pressure from the
water is normal to the face at points of a screw of uniform pitch

working with a slip s, the efficiency of each element and of the

whole screw will be i s. Since the friction must reduce efficiency
in all cases, it follows that upon the above supposition the efficiency
of a screw cannot ever exceed often thought that it
i s. It is

is mechanically impossible for the efficiency of a screw to exceed

i s. This, however, is not necessarily so. This limitation is

associated with and dependent upon the assumption that the

resultant pressure at each point of a screw surface is perpendicular
to the surface. If the water can be made to move in such a man-
ner that the resultant reaction is at an angle with the normal to
the blade surface, we may have an efficiency, neglecting friction,

greater than i s. This is an important point and worthy of

careful investigation.

Referring to Fig. 166, suppose we have acting on a point two

forces OA and OB whose resultant OC makes an angle a with the
axis of x, as indicated. Let the point O be moving with the
velocity OE at the angle P with the axis of x as indicated. Then
the work OA OA X OD.
done by the reaction against the force =
The work done by the force OB = OB X ED = AC X ED.
Draw OF perpendicular to OC and denote EOF by 7. The
ratio between the work done by the force OB and the work done

by the reaction against OA is


Now j3
= 90 a j.
The above is readily applied to the propeller problem. Refer-

ring to Fig. 167, which partially reproduces Fig. 166, consider an

element at O whose pitch angle is denoted by 6. DOP
OC the resultant reaction upon the element O. Draw OF per-

pendicular to OC. Then AO is the transverse force upon the

element denoted by q, say, while is the thrust denoted AC
by /.

OD is transverse velocity V and DE is velocity of advance V&.


POD being the pitch angle POE is the slip angle $. Then the

efficiency is the ratio between the useful work done by t and the
T) 7?
gross input or work done by q and as before is - - Now if DE
DP is speed of screw and PE = =
is speed of advance -= slip ratio s.

The efficiency of the element depends upon the directions of the

resultant OC, and OF
the perpendicular to it. Suppose the re-
sultant OC is perpendicular to OP, then 7
= <, F goes to P and

the efficiency is = i s. It appears, then, to be rigidly

demonstrable that the resultant reaction at every point of a


true screw is perpendicular to the face the efficiency of every ele-

ment, and hence of the screw as a whole, is i s. As the direc-
tion of the resultant OC approaches the fore and aft line, or the
perpendicular to AD, the efficiency of the element increases and
would become unity if the resultant could become perpendicular
to AD. As the swings out from the
direction of the resultant OC
fore and aft line beyond the perpendicular to the element, the effi-
ciency becomes less than 1 5. Friction and head resistance

always tend to swing the resultant in this direction, and the smaller
the slip the smaller the values of AO and OC and the greater the
relative effect of the force due to friction and head resistance.

now, neglecting friction at first, develop the formulae for

I will
thrust and torque of a screw, following the three theories already
referred to. For convenient comparison a uniform notation will
be used, so far as practicable, differing slightly from the several
notations of the original authors.
Rankine's Theory of Propeller Action.
4. Referring to Fig.
165 by Rankine's theory, considering the annular ring of mean
radius r,

Annular area = 2 irrdr.

Volume of water acted on per second =

2 irrdr X AE = 2 irrdr X ^ (i
- 2
5 sin 6).

Stern ward velocity communicated = EC = s -*- cos 6

2 = s^ cos 2 6.
Hence elementary thrust = mass of water per second X stern-

ward velocity imparted = dT =-2 -n-rdr^ (is sin 2

6} s
cos 2
60 60

= w * s (i s sin 0) cos
2 2
g 3 6o

Let q
= cot = Then 2 Trrdr = " ,

dq. sin
2 = -
p 2 TT i + q
cos 6 =-

w 2 R2 2
dT = ~ pA 5 ( Q
+ q M/ 2ir

g 3600 \i + q

^L qdq

At the = o. Then, neglecting the hub, which a very slight

axis q

investigation shows to have very little effect; if q denote now cotan-

gent of the pitch angle of the blade tips, we have on integrating

the expression for dT:

_ _ loge(i + 2
<7 ) _
g 3600 2 7T _2 2 \ 2 21
W />
^ r iog.(i +^ 2
) /io gc (i +9 2
) i

g 3600 47rL ^
\^ 2

i+ q
Now pq = 2 irr p
2 = 4 7i
2 2
r *-*- = irr
2 = if d is extreme
47r 4
diameter. Whence
w /log, i
g 3600 4 L q
\ q
i + q

Whence finally

T = _jrw_ p2d 2
R2 J I _ loge (i + q
) _ s
/loge (i + q
} i_Yl
14400 g L q
\ q
i + 2
q /]
And the torque Q= 2 7T
5. W. Froude's Theory of Propeller Action. Consider now
Fronde's theory.
If the total blade length of all blades at radius r, then the
/ is

total elementary plane area at this radius is Idr. This area ad-
vances at the angle (j> (Fig. 165), with velocity OC, and from
Froude's experiments if a is a thrustcoefficient, we have a result-

ing pressure normal to the blade = Idr aOC sin <. The ele-
mentary thrust is equal to this pressure X cos 6.
Then dT = Idr aOC' sin cos 6. <


cos 6.

Also cos2 6 =


-afcx)' *: ddi + f'

Whence, neglecting the hub as before,

3600 /oOI+^ 27T

The quantity under the integral sign is evidently dependent only

on shape and proportions of the propeller and independent of its
dimensions. It can be determined in any case by graphic integra-

tion. For the present, let us denote it by the symbol X. Then

from Froude's theory T= -
-p*R*dsX, and as before O= -
3000 2 TT

6. Greenhill's Theory of Propeller Action. Coming finally to

Greenhill's theory, we have (Fig. 165)

Elementary area = 2 irrdr.

Velocity of feed of the water

= AC = *-
(i s)
= *-
(i s).
2 7T 6O

= = *
Transverse velocity s *-
cot 6 su>r = s r.
2 7T 60

Transverse momentum per second = - 2 irrdr

(i s) s
^ r
g 60

= Wp R
, x
, 2
s(i s) 4 irrdr.
g 3600
Torque = transverse momentum X r.
7j TV
Whence dO = - p s (i s) 4 ir^dr.
g 3600
- R
_,_, =
2 irdO
s (i
8 3 rdr.
p g 3600

Integrating from r
= o to r = - we have

IV R / \ TT iV / \

g 3600 28800 g

And as before Q = ^2 7T

In connection with Greenhill's theory, it should be pointed out

that the excess pressure at any radius is very simply expressed.
w R2
We have above dT = - s(i s) S-n^rdr.
g 3600
But if AP be the excess pressure per unit area, dT = 2 wrdrAP.
w R 2

Whence dividing through AP = -

s (i s)
2 2
4wr .

g 3600
In other words, the excess of pressure varies as the square of the
radial distancefrom the axis.
7. Comparison of Theories with Each Other. Now, com-
paring the three formulas for thrust and torque, it is seen that
each one is composed of a coefficient, of a term involving the

dimensions and revolutions or speed, and of a term varying with

shape, proportions and slip but independent of the dimensions.
Assuming, as is evidently possible, that we can expand in the X
formula from Froude's theory in the form a fts + negligible

terms, we can write for each formula T = (pdR) (as 13s-). <j>

For Froude's theory (pdR)

= pz dR- and for Rankine's theory

(pdR) is d R of the
z 4 2
(pdR) is p~(PR For Greenhill's theory
<f> ,

same dimensions as before but independent of the pitch. Now,

considering a and /3, it is evident that by the formula for Froude's
theory be very small indeed compared with a. In the
/3 will

Rankine theory formula /3 will be smaller than a, but relatively

larger than in the Froude theory formula. In the Greenhill theory
formula /3 always. = a
neglecting friction, we would have on the theory of all

motion communicated to the water perpendicular to the blade

Q = PI = - - /3s ). 2
<f>(pdR) (as
2 IT 2 7T

As a matter of fact, a very brief examination of experimental

results shows that this cannot hold. If it were true, we could
never have an efficiency greater than i s, and even when fric-
tion is considered we
get experimental efficiencies greater than
i s. So it appears well to adopt tentatively as the general
expression for the torque Q= 2 7T
$ (pdR) (ys 8s

8. Friction and Head Resistance. Now consider friction and

head resistance. Referring to Fig. 165, if / denote the' coefficient
of frictionand dA an elementary area, we have with close approxi-
mation = fdAOB?. In practice
frictional resistance is a much <

smaller angle than indicated in Fig. 165. where it is exaggerated

for clearness. Suppose / is large enough to cover all edgewise
resistance skin friction and head resistance together.

Then dA = Idr, OB* = p*R* cosec2 P = p 2 R 2 (i+q2), q

= >

2 IT

F = f% -l
(i + q>) = ffdR* -*- \ (i +? 2
) dq I
2 IT 2 IT (a }

Fore and aft component = Deduction from thrust

= F sin 6 = fp
dR2 -- -
Vi+q*dq=dTf .

2 TT a

Transverse component =F cos 6 = fpzdR2 -'-q\/i-}-q 2

2 d TT

Difference of torque =F cos 6 X r =F cos 6 "

2 7T



Deduction from thrust for friction

2 ird

Addition to torque for friction = Q,=

2 7T / 2 TTtt

= --fp*dR 2Z, where Z= f

J 2 ird ^
- 2
V T+~fdq.

2 T

Since for the working portions of actual propellers <?

is greater
than i, we will have in practice Z much greater than Y, and it is

reasonable to ascribe the total friction loss to increase of torque.

If we assume - constant = mean width ratio X number of blades,

we can readily determine a curve of Z on q by plotting a curve of

^2^/j 02
*- and integrating graphically.
2 7T

For actual propellers Y and Z can be determined without

on q curves of -
7 __
Vi + o


Vi + o

culty by plotting
2 Tra 2

and integrating graphically.

Fig. 1 68 shows curves of Y and Z and of for elliptical blades

with hub diameter .2 the extreme diameter, plotted upon pitch

ratio, and Fig. 169 shows curves of X
for various values of s,

namely, s = o, .20, and .40.


9. Final Formulae on Theories of Rankine, Froude and Green-

hill. Then the final formulae for thrust and torque including the
friction term can be expressed in the forms below:

Rankine's Theory: T= pWR 2

- /3s )
2 - fdj?R 2

Q = -
- 5s ) + fdp*R Z],
2 2

2 7T

Froude's Theory : T = p3 dR2 (as - /3s )

2 - fdf&Y,

Q = 2-- [p
dR 2 (ys
- 5s
) +fdp R Z].
3 2


Greenhill's Theory: T = d*R2 (as

- /3s )
-fdp R Y,
3 2

Q = ^~ [d*R (ys
2 - 3s
) +fdp R Z].
3 2

2 7T

The above equations are simply to show the form of the ex-

pressions. They do not imply that a and /3 in the Rankine Theory

equation will be the same as in the Froude or Greenhill Theory

equation, but simply that in each case a and /3 will be constant for
a given propeller. The actual values of the constants will vary
with the theory used.
The formulae on Froude's theory are expressed in the above
form, as previously noted, by assuming that can be expanded X
with sufficient approximation in the form C sD, where C and
D are independent of s. It is evident from Fig. 169 that this can

be done and that D

is much smaller than C.

In the theories, as has already been pointed out, it is assumed


that the net reaction at each point is perpendicular to the blade

surface. If this were true, we would always have a = y, /3
= <5,

and the efficiency could never exceed i s even if there were no-
friction. Since experience shows this is not the case, and as from

considering the probable motion of a particle of water it is evi-

dently not necessary that the net momentum impressed upon it
shall be perpendicular to the blade surface, I have, while follow-

ing the same form, used

different coefficients for the torque

expression, expecting that these coefficients y and 5 need not

necessarily be the same as a and /3 used for thrust.

It seems difficult at first sight to conceive of any fluid action


upon a frictionless surface that is not at right angles to it, but if

we consider the matter from the point of view of the velocity im-

pressed upon the water the difficulty disappears. The suction of

the propeller upon the water ahead of it causes a velocity which is

practically all axial, or in the direction perpendicular to the plane

of the propeller disc. Hence, the reaction upon the water is partly
axial before the water reaches the propeller disc and partly normal
or nearly so as the water passes through the disc, the final result-
ant being at an angle with the normal in the direction which we
have seen tends to make the efficiency greater than i s.

10. Comparison of Theories with Facts of Experience. It

does not require much reflection to render it evident that none of
the three theories considered correctly represents the physical

phenomena. This conclusion is very strongly confirmed by the

results ofmodel experiment and general experience.
On Rankine's theory the water while passing through the screw
disc is given the stern ward velocity EC (Fig. 165). This can
occur only if the stream contracts materially while passing through
the propeller or if a material quantity of water from abreast the
disc always flowing into it. Neither motion seems reasonable.

Furthermore, on Rankine's theory, the thrust and torque are

independent of the blade surface, one assumption of Rankine's
theory being that "the length of the screw and number of its
blades are supposed to be adjusted by the rules deduced from

practical experience, so that the whole cylinder of water in which

the screw revolves shall form a stream flowing aft."
Practical experience withmodel propellers shows clearly that the
result assumed by Rankine is unattainable. Rankine's theory
further ignores variations of pressure which must occur in pro-

peller action.
Froude's theory goes to the opposite extreme of Rankine's. It
assumes that the thrust increases always in direct ratio to the
area. Model experiments show conclusively that, while within

practicable limits thrust does increase as long as area increases,

the increase in thrust is by no means proportional to the area

increase, the rate of increase with area diminishing steadily as

area increases.

Greenhill's theory has the same obvious defect as Rankine's, in

that it neglects the effect of area of blade. The portion I have
used ignores the sternward velocity, deducing thrust entirely from
the pressure set up by rotating the water in the disc, but it should
be pointed out that his 1888 paper gives some consideration to
other possible motions involving axial velocity of slip in the
As, then, it seems that no theory we have considered can exactly
represent the action of propellers, it would be necessary, in case we
wished to adhere to formulae, to compare each formula with experi-
mental results and select that one which seemed to agree most
closely. Then using this as a semi-empirical formula, with coeffi-

cientsand constants deduced from experiments or experience,

problems could be satisfactorily dealt with. But it will be ob-
served that each formula of the proper dimensions to satisfy

the Law of Comparison. Hence if either formula holds, the Law

of Comparison will hold, and experimental results, instead of
being utilized to supply coefficients and constants for use with a
formula, can be reduced to a form to be utilized directly by graphic
methods. Per contra, if the Law of Comparison does not hold, the
formulae on all of the three theories will fail. In either case there
isobviously no advantage from a practical point of view in attempt-
ing to reduce the formulae to forms for use in practice. A serious
practical disadvantage is the fact that the formulas use a true slip,
based upon true pitch, or a blade of no thickness. The face pitch
of a blade with thickness, or its nominal pitch as it may con-
veniently be called, is very different from the virtual or effec-
tive pitch, and this fact causes material complications in using

Slip Angle Values.

ii. In connection with theories of pro-
action it is desired to invite particular attention to the fact
that propellers in practice operate with slip angles that are very
small indeed. A slip of 20 per cent somehow seems to imply a
it usually means in practice an
but as a matter of fact
large angle,
angle of from i\ to
5 degrees only, and most propellers show their
maximum efficiency at slips below 20 per cent.

Referring to Fig. 165, where <p denotes the slip angle, we have

sin <> = CD BCcose

CO \/ r\ /(
i A/^~

^ cos 6
2 7T

2 2
r +,


Let y denote diameter ratio =- = Then r = "-

p p 2

Substituting, clearing and reducing, we have finally

sin =5
+ / VVy +
2 2
- S)

Hence given s and y the value of is fixed. </>

Fig. 170 shows graphically the relation between slip angle (j>,

slip 5 and diameter ratio. Also at the top of the figure is a scale
for pitch ratio, but reference to diameter ratio is more illuminating.
Considering a screw of uniform face pitch it is seen that for a given

slip per cent the slip angle is a minimum where the diameter ratio
isgreatest at the blade tip. As we go in from the tip the slip
angle increases, reaching a maximum when diameter ratio = .3
about, and then rapidly decreasing to zero at the axis. But on
account of the hub the falling off of slip angle below diameter ratio
of .3 is immaterial, and to all intents and purposes slip angle in-
creases from tip to hub. The actual values for the diameter
ratios and slips found in practice say below diameter ratio of i.i

and slip ratio of .30 are quite small.

The maximum efficiency of most propellers corresponds to a
nominal the neighborhood of 15 per cent, and for this the
slip in
maximum slip angle at the hub is less than 5 and for the most
important part of the blade it is in the vicinity of 3. These are
small angles, and the fact that slip angles are so small should never
be lost sight of in considering operation of propellers.

21. Law of Comparison Applied to Propellers

i. Formulae for Applying Law of Comparison to

In connection with the Law of Comparison the formulae for the
application of the law to propellers have been already indicated,
but they are recapitulated below.
Suppose we have a propeller and a smaller similar propeller or
model. Let us use symbols as in the table following :

of value in the investigation of propellers. But upon con-


sideration it is evident that in each case the atmospheric pressure

is transmitted through the water, appearing both in front of and

behind model and propeller; and, since the forces upon model
and propeller are due to reactions caused by the motions impressed
upon the water, the Law Comparison will apply provided the
motions of the water around model and propeller are similar.
The pressure relation fails in precisely the same way in passing from
models to ships, but in this case the motions produced are not
affected by the surface pressure and the Law of Comparison holds.
Hence we may rely upon the Law of Comparison and design pro-
pellers upon the basis of model results if we can but be sure that
the motions of the water around model and propeller will be

Now, weare reasonably certain that until we reach speeds and

thrusts at which the phenomenon known as cavitation makes its

appearance the motions of the water around model and propeller

are so nearly similar that the Law of
Comparison is applicable.
When cavitation is present the Law of
Comparison fails, because,
as will be seen when discussing cavitation, the model does not cavi-
tate as a rule, and hence results from it are an unsafe guide when

dealing with the full-sized screw. But the majority of propellers

as fitted are not very seriously, if at all, interfered with by cavi-

tation, and for such propellers model experiments are of great

value, since the Law of Comparison may be somewhat confidently
reliedupon in connection Exact comparison of experi-
with them.
mental data from a model and a full-sized propeller of large dimen-
sions has never been made, but experiments at the United States
Model Basin showed that for small or model propellers ranging
from 8 inches to 24 inches in diameter the Law of Comparison
applies reasonably well. (See paper entitled "Model Basin Glean-
ings," Transactions Society of Naval Architects and Marine Engi-
neers for 1906.)
Wehave seen that theoretical formulas for propeller action all
give the result that for a given propeller form advancing with a

given slip the thrust and torque vary as the square of the speed of
advance and also, that the thrust varies as the square and the
torque as the cube of the linear dimensions.

If this is the case, the Law of

Comparison necessarily holds.
There are a number of reasons for thinking that thrust and torque
for a given propeller advancing with given slip vary as the square
of the speed of advance. If the lines of flow or paths followed by
the particles of water are the same, whatever the speed, then
thrust and torque must vary as the square of the speed. For then
the quantity of water acted upon must vary directly as the speed,
and the velocity communicated to each particle acted on must
vary directly as the speed. Hence the momentum generated per
second, to which thrust and torque are proportional, must vary as
the square of the speed.

Experiments made at the United States Model Basin in 1904

with 1 6-inch model propellers between speeds of three and seven
knots showed that within the limits of experimental error thrust
and torque varied very approximately as the square of the speed.
The propellers whose thrust varied as a greater power of the speed
than the square were usually those with very narrow blades.
Those whose thrust varied as a lesser power of the speed than the
square were usually those with very broad blades.
Finally, experience in analyzing accurate trial results shows that,
broadly speaking, when cavitation is not present, at speeds where
the resistance of the ship is varying as the square of the speed the
slip is practically constant, which of course means that the thrust
of the propeller advancing with this constant slip varies as the

square of the speed.

At speeds for which the resistance of the ship is varying as a
less power of the speed than the square the slip is falling off, and
at speeds for which the resistance is varying as a greater power
of the speed than the square the slip is increasing. This is
fairly strong evidence from accumulated experience that the
thrust of full-sized propellers varies as the square of the speed of
In the light of present knowledge we appear to be warranted in
concluding that the Law of Comparison applies to propeller action
sufficiently well for practicalpurposes until cavitation appears.
There reason to believe, however, that cases have occurred

where cavitation has been present without being suspected.


22. Ideal Propeller Efficiency

i. Thrust, Power and Efficiency of Ideal Propelling Apparatus.

In a paper before the Society of Naval Architects and Marine
Engineers, in 1906, entitled "The Limit of Propeller Efficiency,"
Assistant Naval Constructor W. McEntee, without setting up

any special theory of propeller action, has pointed out the limit of
propeller efficiencybeyond which we cannot go.
Suppose we have a frictionless propelling apparatus discharg-
ing a column of water of A square feet area directly aft with an
absolute velocity u, while the speed of the ship is v, both v and u

being measured in feet per second. Then if w denote the weight

per cubic foot of the water, the weight acted on per second is
wA + u} and the mass is
-A +
(v ( ).
The reaction or thrust T= A (v + u) u being equal to the
sternward momentum generated per second.

Useful work = -A 1$)

(v + u} vu.
There being no friction, the lost work is simply the kinetic energy
in thewater discharged. Hence we have

Lost work =-A (v + u)

g 2
fiat not J


Gross work (v + u) vu -\ A (v + u) 2
g g
Useful work v
Efficiency e
= -77 r- =
Gross work . u
v H

Also solving for u in the equation for thrust T, we get

/v sT v
u =

V 4 wA 2

Substituting in the expression for efficiency, we have

- -

This expression for maximum efficiency must involve the assump-


tion that the water is discharged without increase of pressure.

The effect of an increase of pressure would be to decrease the
efficiency, since work done against pressure would be done with

- v Hence we conclude that the value

efficiency of e above
v + u
is the maximum that could be attained by a perfect propeller.
Suppose, applying this to a screw propeller, we write for A ,

where d the diameter of the propeller in feet. Now if
is denote U
useful horse-power delivered by the propeller and P denote gross

horse-power, or horse-power delivered to the propeller, we have

TV ~ M eP 6080
whence T =
eP = rr =
U ,
Also v = - - T where V

55 v 3 6o
is speed of advance in knots. And g
= 32.16, w = 64 for sea
water. Substituting and reducing, we have finally

16 24 e + 8 e
_ 2 3 e + e

s 3
d?V 292.2 e 36.52

Discussion of Ideal Efficiency Results.

2. From the above,
Figs. 171 and 172 were drawn, Fig. 171 showing contours of
V and
efficiency on values of as abscissae of as ordinates and
Fig. 172 showing contours of efficiency on values of d as abscissae

and of 7-3 as ordinates.

These figures should not be mistaken as representing actual
efficiencies that are attainable. They are purely ideal diagrams,
and their indication that efficiency always increases with increase
of diameter is if followed too far as regards actual pro-
pellers. They are interesting and instructive, however, as giving
us in any particular case a limiting efficiency beyond which we
could not possibly go and which we must fall short of in practice.
In Fig. 171 there is shown a supplementary scale of or power
A ,

per square foot of disc area. This of course bears a constant

ratio to

A striking result of the formula for ideal propeller efficiency is

the high efficiency attained with large slips. The expression for
terms of and The formula

- Si=
slip ratio s\. in v u, is for
v + u
efficiency is e
= ,
whence expressing e in terms of s\, we have

Fig. 173 shows a curve of e plotted on Si, as deduced from the

above formula. This efficiency is everywhere above the line i Si.
In this connection it is interesting to recall that numerous ex-
periments with model propellers at high slips show an efficiency
greater than i s. It should be remembered, however, that in

the case of these actual small propellers s is derived from the pitch
of the driving face, while in the ideal formula $1 is based upon the
assumed sternward velocity u of the water, and the water is not
supposed to have any transverse velocity. The actual sternward
velocity of the water in the operation of actual propellers is not easy
to determine or estimate, and transverse velocity
always present. is

OnRankine's theory we can readily establish the relation be-

tween s and sternward velocity. In Fig. 165 the sternward velocity
is EC = s cos 6. This is much less than BC, the slip velocity.

While we cannot say that in actual cases the sternward velocity is

EC, there is no question that it is very much less than BC, the slip
velocity. It could be equal to BC only if there were no trans-
verse velocity communicated to the water, and there is no ques-
tion that in practice transverse velocity is always communicated.
A very common mistake to consider the sternward velocity

communicated to the water the same as the slip velocity, or in BC

Fig. 165.

23. Model Experiments Methods and Plotting Results

Experimental Propeller Models and Testing Methods.

i. -
Having concluded that the Law of Comparison is applicable to
many cases of propeller action so that experiments with model
propellers may be expected to be of value, I will now go into this

question. Numerous experiments with model propellers have

been made at the United States Model Basin. The details of the
apparatus and methods used will be found in the author's paper
of 1904 before the Society of Naval Architects and Marine Engi-
neers entitled "Some Recent Experiments at the United States
Model Basin." The experimental gear
described in that paper
has been changed subsequently only in minor details as improve-
ments suggested themselves.
The model propellers are usually made of composition, accu-
rately finished to scale. Most of them have been 16 inches in
diameter. When being tested the model propeller is attached to
a horizontal shaft projecting ahead of a small boat which is rigidly
secured to the carriage traversing the basin. The shaft projects
so far that the propeller is practically unaffected by the presence
of the following boat. The
propeller shaft center is 16 inches
below the surface of the water, so that the blade tips of a 1 6-inch
model are immersed 8 inches, or one-half of a diameter. The hub
is fitted with fair- waters in front and behind. Fig. 174 shows the
arrangement for a hub 3! inches in diameter, which was a stand-
ard hub diameter adopted for all models which did not represent
actual propellers. For models of actual propellers, the hubs rep-
resent to scale the actual hubs, appropriate fair-waters being fitted.

Dynamometric apparatus, described in detail in the paper

above referred to, enabled the torque and thrust of the model
propeller to be accurately determined.
By making runs with dummy hubs having no blades attached
the hub effect was eliminated as far as possible, the endeavor
being to determine experimentally the torque and thrust of the
blades alone.
The greater number of experiments were made at a 5-knot speed
of carriage, this speed of advance being kept constant as nearly
as possible, and being varied by varying the revolutions of the

propeller. In the early stages of the experiments, however, a

number of propellers were tested at speeds of advance ranging
from 3 knots to 7 knots, and between these speeds it was found that
within the limits of error the thrust and torque at constant slip
varied practically as the square of the speed. As has been already

pointed out, this agrees with the formula of Rankine, Froude and
Greenhill, which agree in making thrust and torque vary as R 2

and when constant the speed of advance varies as R.

slip is
In making the 5-knot experiments speeds of individual runs of a
series would differ slightly from 5 knots, and the thrust and torque
were reduced to the 5-knot speed by taking them to vary as the
square of the speed.
2. Methods
Recording Experimental Results.
of As will be
seen upon consulting the original paper, during a run the thrust
and torque are recorded continuously, and after uniform condi-
tions have been reached the time and revolutions are recorded
every 32 feet. For convenience the thrust and torque at 5 knots
speed are plotted initially upon the revolutions made by the pro-
peller upon a 64-foot interval denoted by pi, which is one of
the quantities observed.

Fig. 176 shows curves of thrust and torque plotted thus for the
model propeller whose developed blade outline and blade sections
are shown in Fig. 175. This is a 1 6-inch three-bladed model

propeller of the true screw, ordinary type, the pitch being 16

inches pitch ratio i.oo the blades being elliptical, of .25
mean width ratio, and the sections ogival. The hub diameter is

.2 the propeller diameter. The curves of Fig. 176 are plotted upon
Pi, or revolutions per 64-foot interval. Lines showing the values
of pi for various values of the slip are shown on the figure, the

slip being based upon the nominal pitch of 16.0 inches. These
lines are not equally spaced, for, p denoting pitch in feet and s

the we have ppi

= 64, or pi= -
slip ratio, (i s)
P (i
~ s)

equal increments of s the interval between successive correspond-

ing values of pi constantly increases.
It will be observed that p\ is dependent upon the pitch and
slip only and for a given slip is quite
independent of the speed.
Furthermore, the experimental apparatus was such that pi was
determined with great accuracy. Thus it was a very suitable
quantity to use as a primary variable upon which to plot the
experimental values of thrust and torque for the purpose of de-
ducing curves of the same.

24. Model Propeller Experiments Analysis of Results

1. Methods of Plotting Information Derived from Experiment.
- The results of model experiments having been plotted as curves
of thrust and torque upon the revolutions made upon a 64-foot

length as shown in Fig. 176, the lines for various definite nominal
slips being indicated upon the same diagram, the subsequent
treatment depends upon the purpose in view.
For purposes of analysis, comparison of efficiency, etc., the
methods would naturally differ from those most convenient for
use in design.
When we consider the best method of plotting for purposes of
analysis, etc., curves deduced from model propeller experiments, it
soon becomes evident that we may with advantage record the
data as curves of coefficients quantities that do not vary with
dimensions. As abscissae for such curves the slip ratio is a de-
sirable quantity to use. It is not dependent upon size or speed,

and isone of the primary variables involved in screw action.

2. Virtual and Nominal Pitch and Slip. The question at once
however, whether we should use nominal slip, namely, slip
based upon the pitch of the screw face, or real slip, i.e., slip based
upon the virtual pitch, or pitch of the ideal blade of no thickness
which would act as the actual blade.
is a thing very different from the nominal
This virtual pitch
pitch. The
ignoring of this fact has had a great deal to do with

the prevention of correct conclusions as to propeller performance.

In the case of a true screw the pitch of the driving face is known,
but every point of the back has a pitch, and the back has much to
do with screw performance. One might think without looking
into it that for ordinary cases the pitch of the back is nearly the
same as that of the face. The truth is that the pitch of the back
varies prodigiously from the pitch of the face. Fig. 175 shows
blade sections of a screw of not unusual blade thickness and of
face pitch equal to diameter, the sections being of the usual ogival

type. Taking face pitch and diameter as 16 feet, Fig. 177 shows
plotted on radius the pitch of the back at the leading edge and at
the following edge. It is seen that the pitch of the leading por-

tion of the back will average somewhere about 50 per cent less
than the uniform pitch of the face or the nominal pitch. On the
other hand, the pitch of the following edge of the back is on the
average somewhat more than 50 per cent greater than the nomi-
nal pitch. It is quite obvious that such a screw cannot act as a
theoretical screw, having blades of no thickness and of the uniform

pitch of the face. It is evidently desirable to find some method

of determining for a known screw virtual pitch, or equivalent

uniform pitch. Now, for all formulae we have, neglecting fric-

tion, no thrust or torque at zero slip. Experimental results with

screws of uniform nominal pitch and ogival type of blade section
always show as in Fig. 176 both thrust and torque when the slip
calculated on the nominal pitch is zero. It follows that for such
screws the virtual pitch is greater than the nominal pitch. This
might be inferred, too, from the fact that at the rear of the blade
the pitch of the back is always greater than the nominal, and, if
the back has any influence at all, it must increase the virtual

pitch over the nominal pitch. Suppose, now, we consider some

experimental results. Fig. 178 shows upon an enlarged scale the
lower part of Fig. 176, being curves of thrust and torque as deter-
mined experimentally for a 1 6-inch model of the propeller of
Fig. 175 plotted upon p\, or revolutions required to traverse a
distance of 64 feet, the speed of advance of the propeller being

kept constant at 5 knots. Now on any theory we have at true

zero slip a negative thrust Tf and a positive torque Qf, both being
due to the friction and head resistance only. From the formulae

given when considering the theories of Rankine, Froude and


T, = 3
-fdp R*Y, Qf=-- fdp*R Z.

2 7T

Whence fdj?R*=
~^ = ^'
^ f = Y = pT ,
^ when
Whence '
s o.
2 irQf Z 2 -nQ

Now is a fixed quantity for the propeller. For the propeller in

question it is equal to .236, from Fig. 168. Fig. 178 shows the

method to be followed. If s = o, we have p = -

So we can

plot a curve of p on pi as abscissa. Also we can plot the curve of

-^ = Y
;, as shown. This has the value .236 at PI= 42.11,
2 irQ Z
for which p = 1.520 feet. Then from the diagram the virtual

pitch of the screw is 1.520 feet, or 18.24 inches, or 1.140 times the
nominal pitch of 16 inches. Very frequently the virtual pitch is
taken such that zero slip will give zero thrust. This is not quite
correct, however, because at zero thrust there is a small negative
thrust due to friction and an equal and opposite positive thrust
due to slip. The error, however, is not great. In Fig. 178, at zero
thrust PI = 42.70, p = 1.499 feet
= 17.99 inches. The difference in
virtual pitch is per cent, and as it is very difficult
only about 1-3

to make model propeller experiments with minute accuracy, it is

hardly worth while in practice to use the exact method. More-

over, whilewe should always bear in mind that the nominal pitch
isnot the real pitch or virtual pitch, it is very desirable to use
always the nominal pitch in practical cases. We shall see that
this can be done, so that the question of virtual pitch, though of
great scientific interest, is academic rather than practical. So,
except for special applications, results for true screws of uniform
face pitch will be plotted upon nominal slip corresponding to the
face or nominal pitch.
Determination of Efficiency.
3. The ordinates for the curve of

efficiency plotted upon nominal slip are readily and simply de-
termined from the curves of thrust T in pounds and torque Q in
pound-feet. For if p denote pitch in feet, R revolutions per min-
ute and s the slip, speed of advance is p (i s) R, and useful work
done in a minute = TpR (i s). The gross work, or work de-
livered to the model, is Q X 2 irR.
T R ~s
Now efficiency = (Useful Work) H- (Gross Work) = P ^ ">

2 (JirK

Tp(i -s}
Q **

Note that the quantity p (i s) is the advance of the screw


for one revolution, and its value is the same whether nominal or
virtual pitch is used, the slip in each case being that appropriate
to the pitch. Since at the Model Basin the curves of T and Q are

plotted upon pi, the revolutions per 64-foot interval, it is conven-

ient to use this in the efficiency formula.

We have = '

r Substituting and reducing, we have finally


- s)

_ 10.186 T
The values of T, Q and p being taken off for the values of pi for
the various slips, as indicated in Fig. 176, the efficiencies are

readily calculated and plotted on slip.

4. Characteristic Coefficients. The next question is as to the

curves of coefficients which will completely characterize the pro-
peller. Various coefficients may be used. Papers by the author
and Messrs. Curtis and Hewins of the Model Basin staff before
the Society of Naval Architects and Marine Engineers give vari-
ous forms of coefficients, but it is believed that those given below
are simple and convenient.
We have to deal with the power absorbed or propeller power P,
the useful or net power E, the speed of advance in knots V, the
revolutions per minute R, the slip and the size.
Whatever formula we use we are led to the same type of ex-
pression connecting power absorbed, speed of advance and diame-
ter. Thus using Rankine's formula,

Q= 2 7T

Gross power P= = - 1 z
(p RW(ys
- 6s
) + fp*R pdZ).

33000 33000

pR^ ^^-.
s i
m pd=-
J- P
f - ,
(i-s) X330oo m (i -s) X 33000J \7v./
3 / 3

Using either Froude's or Greenhill's formula we are led to the

same expression except that the Froude theory formula will have
the first term in the parentheses divided by m and the Greenhill

theory formula will have it multiplied by m 2

. In either case we
= dV A
2 s

may write P . where A is a coefficient independent of the

sizeand speed of the screw but varying with the slip and depend-
ent upon shape, proportions, etc. The divisor 1000 is introduced
simply in order to give A a greater value than unity in practical
cases. Otherwise A would be inconveniently small. Evidently
then a curve of A plotted on slip will completely characterize a
screw as regards the important question of its capacity to absorb
If E denote the useful or effective horse-power delivered by the

screw, we have E= eP = (PV3

eA --
Let us denote eA by B. Then curves of e, A and B plotted
upon slip will completely characterize the action of a propeller of
given features independently of size and speed.
We have already seen how to determine e from the curves of
Q and T. These curves are

advance and at any given point P -

for a fixed diameter

and speed of

33OOO p (l S) IOOO

Whence A= 2^X101.33
33000 p(i -s)
From the experimental results for a model propeller for a given
value of s we know everything on the right-hand side of the equa-
tion and hence can determine A without difficulty. Similarly, it

will be found that we may derive B from the thrust 7\


B = -
= TV X


Then curves of A, B and e completely characterize a propeller.

<PV3 B


As a matter of fact any one of them can be derived from the


other two. They are all functions of slip and proportions and
characteristics of the propeller and independent of size and speed.
Table XI shows the calculations necessary to determine curves of
A, B and e from the experimental data recorded as in Fig. 176.
shows four curves of A, B and e as deduced from the
Fig. 179
results ofmodel experiments for four propellers of the same nomi-
nal pitch ratio 1.2 and mean width ratio .2 and of the different
blade thickness fractions indicated. The curves are plotted upon
nominal slip and show that for this blade width and pitch ratio
efficiency increases as the blade thickness is reduced, but the power
absorption coefficient A and the thrust coefficient B decrease as
thickness decreases.

Application of Curves of Coefficients from Model Propellers.

- Curves of B are particularly valuable in estimating from model
results the probable performance of propellers of ships. If there

were no reactions between ship and propeller, that is, if the ship
were a "phantom ship" as Froude calls it, which offers resistance
the same as the actual resistance without disturbing the water or

modifying the action of the propeller, the case would be very

For the ship we would know from model experiments the E.H.P.
at any speed V and would also know the diameter d of the pro-

peller. Then B = 3
is known for any speed from consid-

eration of the ship. But from the propeller model experiments

we have a curve of B plotted on slip. So having determined B
IOI ^ if
for a speed V we know what the slip must be. But R= ' "^
- ;

p (i s)
hence we know what the revolutions must be. Finally from the
slip determined by B we may determine e and A corresponding.
We can then determine the power P absorbed by the screw by
either one of the two formulae. We have


We shall see later that the case of the actual ship is

not so simple
as that of the phantom ship, but curves of revolutions and horse-

power deduced entirely from model experiments for phantom ships

agree surprisingly well in many instances with the actual curves
determined by trial of the full-sized ships.
The author has encountered cases where curves of revolutions
and speed obtained from trials of full-sized ships presented fea-
tures which appeared at first sight abnormal but were found to be

duplicated almost exactly by the estimated curves of revolutions

and speed deduced entirely from experiments made independently
with models of ship and propeller. On the other hand, when the
full-sized propeller shows cavitation, the curves deduced from
model results differ materially from the actual curves, a fact
which in some cases permits of the determination with a good
deal of accuracy of the point where cavitation becomes serious.
6. Methods of Plotting Information for Design Work. The
preceding analysis and method of plotting results of model experi-
ments is not very convenient when we come to design work. The
designer of a propeller knows in advance or can estimate with
reasonable accuracy the power P which the propeller is to absorb
and the speed of advance of the propeller through the water VA.
He either knows the revolutions R which are to be used, or, sup-
posing the revolutions may be varied through a certain range,
wishes to ascertain the effect of such variation upon his design.
He then has to determine diameter, pitch, blade area, blade thick-
ness and blade shape.
It is evident, then, that in plotting model experiments for use in

design it would not be advisable to plot them upon slip, because

this not a quantity that is known or can be closely approximated

in advance. It is desirable to use variables independent of size

but involving power, speed and revolutions, etc. There are many
such expressions. For practical applications the following will be
found convenient:

p =

where d is diameter in feet. The quantity p is practically the

same as an expression suggested by Mr. R. E. Froude in discussing
a paper by Barnaby before the Institution of Civil Engineers,

May 6, 1890 (Vol. CII, p. 101). In discussing the same paper


Mr. C. Humphrey Wingfield suggested the use of


The quantities p and 8 may be readily connected with the

coefficients already used in analysis of propeller experiments or
can be deduced directly from the model propeller results.

Thus we have seen that P= d?V3 and we know that
-s) = F
2 2 2
(ioii) .

Multiplying the two together,


Whence ^- = (- -^- )
^^ =
io. 2 68[-
' A
-5)\pl (i \pl (i -5)
= R d
Whence p =3.204
y/ p i s

The right-hand expression for p is independent of size of propeller,

and values of p are correctly calculated from a curve of A plotted

on s. Usually, however, it is just as convenient to calculate them
from the curves of torque, etc., of the model propeller. It will be
found that we may write


Similarly, we may write

(PVA ) * (4)* V^(i


Table XII shows calculations of values of p and 8 for one of the

model propellers whose results are plotted in Fig. 179.
Figure 180 shows for the four propellers of Fig. 179 curves of
efficiency and of 8 plotted on p. The calculations it is seen are
made for various values of s, and on the curves of 8 the spots

corresponding to various values of s are indicated. The scale

used for p a variable one, the abscissa values being proportional

to Vp instead of p directly. This is a convenient device for spac-


ing widely the values of p which are the most important without
extending the p scale unduly.
The application of curves such as those in Fig. 180 to design
work is
very simple.
Thus, suppose a propeller is to be designed to absorb 10,000
horse-power with a speed of advance of 20 knots and to have 200

VA 3200000 320 V Vjf 17.888

From Fig. 1 80 for p = 11.18 the value of 5 for the various blade
thickness ratios varies from 54.8 to 57.6, the corresponding values
of diameter varying from 12.25 to 12.88.
It is is low, only about .66, and
seen, however, that the efficiency
the slip high. Evidently the pitch ratio of 1.2 is not adapted to
the case and should not be used. But suppose the revolutions
desired had been 100. Then we would have
p = 5.59, d = .3551 8.

For this value of p we have good efficiency, and if the law of

comparison holds we would get good results from a propeller of
pitch ratio 1.2. For p = 5.59 the values of 8 range from 54.2 to
58.2 and of d from 19.25 to 20.66. In practice we would choose
a value of d corresponding to a blade thickness fraction, then
determine the actual blade thickness necessary, and if the result-
ing blade thickness fraction differed much from that first esti-
mated, a second approximation would be made using the correct
blade thickness fractions.

25. Propeller Features Influencing Action and Efficiency

A number of experiments have been made with 1 6-inch model
propellers at the United States Model Basin. Many of the results
obtained were published in the Transactions of the Society of
Naval Architects and Marine Engineers for 1904, 1905 and 1906.
These results and others not published enable some conclusions
to be drawn positively as regards 1 6-inch propellers and with con-

fidence as regards propellers of ordinary sizes within the limits

where the Law of Comparison is applicable.
i . Number of Blades. number of pro-
There were tried a

pellers with blades identical but number from two

differing in
to six. It was found that efficiency was inversely as the number
of blades; that is, a propeller with two blades was more efficient
than a propeller with three identical blades, that one with three
blades was more efficient than one with four identical blades and
that one with four blades was more efficient than one with six
identical blades.
Also while total thrust and torque increase as number of blades
is increased, the thrust and torque per blade fall off. A three-
bladed propeller at a given slip does not show 50 per cent more
thrust and torque than a two-bladed propeller with identical
blades. Fig. 181 shows approximately for working slips the rela-
tive efficiencies and coefficients for 2-, 3- and 4-bladed propellers
identicalexcept as to the number of blades. The curves are
curves of ratios of the quantities concerned, those for 3 blades
being taken as unity in each case. As we have seen :

_ loco P ''
R _ i OOP E

where d is feet, V is speed of advance in knots and P

diameter in
and E power absorbed and effective power. The subscripts refer
to the number of blades, A^, for instance, denoting the value of
A for 4-bladed propellers. It is seen that the power absorbed,

depending upon the coefficient A varies more nearly as the number

of blades than the useful horse-power depending upon the coeffi-
cient B. The 2-bladed propeller shows slightly greater efficiency
than the 3-bladed, and the 4-bladed distinctly less. It should be
remembered that Fig. 181 refers propellers working under
identical conditions of slip, speed of advance, etc. This means
that a 4-bladed propeller will absorb about 30 per cent more power
than a 3-bladed and a 2-bladed propeller about 15 per cent less.
In practice the question to be decided is whether to use a
4-bladed or a 3-bladed propeller when the same power is to be
absorbed. In this case the 4-bladed propeller would be smaller

than the 3-bladed and hence might have a pitch ratio more favor-
able to efficiency than the pitch ratio of the corresponding 3-bladed

propeller. So the question of 3- or 4-bladed propellers would re-

quire investigation in each case. The methods to be used will be
considered later.
2. Outline or Shape of Blades. The question of shape or
outline of blade faces has been given much attention in connection
with propeller designs and in some cases extravagant claims have
been made for special shapes.
Fig, 182 shows five blade shapes which were experimented with
at the United States Model Basin. Blade thickness fraction was
constant in each case, being .0575. Three pitch ratios were used,
.8, i.o and 1.2.

The results were quite consistent and showed that the blades
with broad tips absorbed more power and gave more thrust but
with slightly less efficiency. While the very pointed blades showed
up slightly the best, there is some reason to doubt whether they
would retain their superiority which was not very marked in
full-sized propellers. The experiments
justify us in looking with
doubt upon claims for great gain of efficiency by reason of some
special shape of blade, and appear to indicate that for all-round
work the old well-known elliptical shape is probably as good as

any, though it may be that some other oval shape may be found
slightly better. On the other hand the conclusion seems warranted
that if circumstances render some special shape desirable, it can

be used without serious loss of efficiency provided it is not alto-

gether abnormal.
3. Rake of Blades. It is "a
very common practice to rake or
incline the blades of a propeller aft. Sometimes they are inclined
forward. At the United States Model Basin, six propellers, all
of .2 mean width
ratio and .0425 blade thickness ratio, were tested.
Three were of .6 pitch ratio and three of 1.2 pitch ratio. Of each
trio, one had the blades inclined 10 aft, one had the blades set
normal to the shaft and one had the blades inclined 10 forward.
The diameter was 16 inches in each case. Fig. 183 shows radial
sections of the blades. The experiments gave almost identical
results, the difference of torque, thrust, and efficiency being

slight. So far as efficiency goes, then, there seems no reason to

rake the blades of propellers. The advantage sometimes claimed
for blades raking aft is that they prevent a supposed centrifugal
motion of the water. Careful investigation of 1 6-inch propellers
on test failed to show any evidences of centrifugal action except
for some models of very thick blades and coarse pitch tested at
3 knots speed of advance with a slip of 75 or 80 per cent. These
models were practically standing still and seemed to throw the
water out under the conditions described. Numerous experiments
with 1 6-inch propellers under normal conditions showed the propeller

race to be practically cylindrical and that so far from there being

centrifugal motion, there is a slight convergence abaft the propeller.

There doubt that the advantages of rake as regards pre-
is little

vention of centrifugal motion are imaginary.

A real advantage of rake in practice is that the blade tips of

side screws are thereby given greater clearance from hulls of usual
form than if the blades were radial or with the same blade clear-
ance strut arms are shorter. A very real disadvantage is the
increase of stresses in the blades because of centrifugal action.
This will be discussed later. It is a serious matter for quick run-
ning screws, and for such screws at least blades should never rake.
4. Size of Hub. One of the features of the Griffith screw
introduced some fifty years ago was a large hub sometimes
with diameter a third that of the propeller. These screws were
often very successful, and as a result of practical experience there
have for many years been advocates of large hubs. Experiments
with model propellers at the United States Model Basin have
shown that large hubs are distinctly prejudicial to efficiency.
Full-scale experiments with turbine vessels seem to have shown
the same thing, material gains in speed having been reported
after substituting solid propellers with small hubs for propellers
with large hubs and detachable blades. The argument against
the large hub is very simple. When a large spherical hub is mov-
ing through the water there must be a strong stream line action
abreast its center, the water flowing aft. Hence the inner por-
tion of the blades must be working in a negative wake produced

by the hub a condition prejudicial to efficiency.


It is sometimes argued that with a small hub the inner portion

of the blades offer more resistance than if they were suppressed
and a large hub fitted.

This probably not true, especially when we consider that the


large hub appreciably increases the vessel's resistance. But even

if it were true, the prejudicial effect of the large hub upon the

blade outside of it would be enough to turn the scale against it.

With slow-running screws of coarse pitch the large hub, while

prejudicial to efficiency, will not affect it seriously; but for screws

of such fine pitch as usually fitted in turbine work the inner parts
of the blades do relatively more work and are relatively more
efficient than in the coarse screws. Hence, reduction of the work
done by them and of their efficiency through a negative wake set
up by a large hub is likely seriously to reduce the efficiency of the
screw as a whole.
5. Standard Series of Model Propellers. have now con- We
sidered the minor factors affecting propeller operation and effi-

ciency and will pass to major factors. These are pitch ratio, blade
and slip. In considering resistance of ships
area, blade thickness
the major factors of residuary resistance were investigated by
means of a standard series of models whose variations covered
the useful range of the major factors concerned. Similarly, the
fieldhas been covered for propellers by a standard series of models
of varying pitch ratio, mean width ratio, and blade thickness frac-
tion. They were all 3-bladed propellers 16 inches in diameter,
with blades that were developed outline. The hubs
elliptical in
were cylindrical and 35 inches in diameter, being practically .2 of
the propeller diameter. Six pitch ratios were used namely, .6,
.8, i.o, 1.2, and 2.0. For each pitch ratio five blade areas were
used. Fig. 184 shows the developed areas of the five blade faces.
Their mean width ratios, as shown, were .15, .20, .25, .30 and .35.
Six pitch ratios and five mean width ratios resulted in 30 propellers.
These were made true screws with ogival blade sections, the backs
being circular arcs, and with extra thick blades.
After being tested, the thickness was reduced by taking metal
off the back to form new ogival sections, the face being untouched,

and thus new propellers with the same faces as before, but thinner

blades, were made. These were tested as before. This process

was repeated twice, so that each blade was tested in four thick-
nesses, being finally unusually thin. This made 120 propellers
tested in all. Table XIII gives their data. The original pro-
pellers are numbered i to 30 and the successive cuts denoted by
the letters A, B and C. Great care was taken when reducing
thickness not to change the face, and toward the edges the recut
blades were probably a shade thicker than true ogival sections.
It is difficult to make model propeller experiments with minute
accuracy, but in this case, owing to the number of propellers tried
and the number independent variables involved, irregular ex-

perimental errors could be practically eliminated by cross fairing

on pitch ratio, mean width ratio and blade thickness fraction.
Figures 185 to 208 show the experimental results after this was
done in the form of curves of thrust in pounds, torque in pound
feet and efficiency. All refer to a 5-knot speed of advance. The
results are plotted upon nominal slip as being most convenient for
practical applications.
The results of trials of these 120 propellers are worthy of the
most careful study. We will now consider them
briefly in con-
nection with the influence of pitch ratio, blade area, blade thick-
ness and slip upon thrust, torque and efficiency.
6. Pitch Ratio. The effect of variation of pitch ratio is illus-
trated in Fig. 209, which shows for propellers of .25 mean width
ratio and .04 blade thickness fraction curves of maximum effi-

ciency and of thrust and torque for 20 per cent slip. This
figure is typical. It is seen that for constant slip and speed of
advance, torque and thrust increase as pitch ratio decreases, the
increase becoming more and more rapid as pitch ratio becomes

The efficiency remains nearly constant over a fairly wide range

of pitch ratio having its greatest value at a pitch ratio of about

1.5. As pitch ratio decreases, however, efficiency begins to fall

off, and below the value
of unity the falling off is rapid. In prac-
tice screws of fine pitch have frequently shown very low efficiency
as a result of cavitation, but apart from this, screws of fine pitch,

say below a pitch ratio of unity, are essentially less efficient than

screws of pitch ratio 1.5 or so, and the smaller the pitch ratio the
less the efficiency.

7. Blade Thickness. When we study the influence of blade

thickness we
find that the thicker the blade the greater the thrust
and torque for a given slip. This is perfectly natural when we
reflect that the results are plotted upon nominal slip and that the
thicker the blade the greater the virtual pitch. The effect of blade
thickness upon efficiency is summarized in Fig. 210. It was found
that for a given blade area the relative variations of efficiency with
blade thickness were nearly the same for slips used in practice
regardless of pitch ratio. Hence Fig. 210 shows for each blade
width an average curve of relative efficiency plotted on thickness
only; for each curve, unity corresponds to a different blade thick-
ness fraction, the broad blades being thinner than the narrow
blades. This is generally in accordance with what considerations
of strength necessitate in practice.

Figure 210 indicates that the efficiency of narrow blades increases

rapidly as they are thinned, while for the broad blades thickness
has little effect upon efficiency, and in fact the thicker blades
seem slightly more efficient. When we remember that on ac-
count of strength a narrow blade must be thicker than a broad
blade the deduction from Fig. 210 is that practicable variations
of blade thickness will have comparatively little effect upon

efficiency. This conclusion, however, is from results of experi-

ments where cavitation was not present, andit is generally agreed

that to avoid cavitation propeller blades should be as thin as

It is probable that in many cases if the blades are made too
thick cavitation would
reduce efficiency without the propeller
actually breaking down, while it will be avoided altogether with
thin blades. Hence we should make propeller blades reasonably
thin in practice, in spite of Fig. 210. Where cavitation is likely
they must be made may be remarked, however, that
thin. It

Fig. 210 appears to be in general accordance with facts of ex-

perience with slow-running propellers. Coarse, heavy propellers of
this type often give very good results in service in spite of thick

8. Blade Areas. In the experiments with the standard series

of propellers was not practicable to investigate the question of

blade area entirely apart from that of blade thickness. The broad
blades were made thinner than the narrow ones, as would be the
case with actual propellers in practice when
a question be-
it is

tween a narrow-bladed propeller and a broad-bladed propeller to

absorb the same power at the same revolutions and speed.
It is owing to the greater thickness of the narrow blades, and

hence their greater virtual pitch for a given nominal pitch, that in
the fine pitches the narrow blades actually absorb more power and
deliver more thrust for a given nominal slip than the broad blades.
In the coarse pitches this is not the case for slips such as occur in
practice, but the broad blades do very little more than the narrow
Even after making allowances for the thickness effect it is evi-
dent that the broad blades by no means absorb torque and deliver
thrust in proportion to their areas. In fact the influence of blade
area upon thrust and torque surprisingly small.

Considering efficiency it is seen that for propellers of pitch ratio

usually found in practice the broad blades and the narrow blades
are both less efficient than blades of medium width, say with a
mean width ratio of .25 to .30. The differences are not great,
however. It is interesting to note the superior efficiency of the
narrow blades for the propellers of
abnormally fine pitch. This,
however, is not due to the fact that the blades are narrow, but to
the fact that the narrow blades have greater virtual pitch ratio,
and for the propellers of very fine pitch gain in virtual pitch ratio
means gain in efficiency.
The experiments with the standard series of model propellers
warrant fully the broad conclusion that, when cavitation is absent,
propellers may vary quite widely in pitch ratio (above 1.2 or so)

and in area with little change in efficiency, provided diameter is

such that they work at slips at or near that of maximum efficiency.
This conclusion is fully borne out by experience, which has led
many people to conclude that there was so little difference be-
tween propellers that any propeller which allowed the engine to
develop its power at the desired revolutions and showed a good

slip was a good enough propeller. For low-speed work this is

reasonably correct; for high-speed work, even leaving out of ques-
tion cavitation, propellers which absorb the power at the desired
revolutions are liable to vary seriously in efficiency, particularly if,
as is usually the case, they must be of the fine pitch type.
9. Figures 185 to 208 show that all curves of efficiency

plotted upon slip present the same general appearance. Con-

sidering nominal slip the efficiency is zero at a certain negative
slip. The thicker and narrower the blade the greater in general
the increase of virtual over nominal pitch, and the greater the
numerical value of the negative slip corresponding to zero effi-
ciency. It will be noted, however, that for the narrow blades of

pitch above unity there seems to be a slight falling of! of virtual

pitch with thickness beyond the A cut. This is probably due to
the fact that as the thickness of these narrow blades is increased
a point is reached where the water breaks away from the back, the
latter losing its grip, as it were. The process is analogous to cavi-
tation, though cavities are not formed. As the slip increases from
that corresponding to zero efficiency, the efficiency rises very

rapidly at first, then reaches a maximum and thereafter falls off.

The nominal slip corresponding to maximum efficiency is nearly
always between 15 and 20 per cent for blade thickness that
would be used in practice, but slip can be increased to 25 per cent,
and in some cases to 30 per cent, without serious loss of efficiency.
But such an increase means an eno mous increase in thrust and
torque. Hence a given propeller will vary widely its power and
thrust without material change of efficiency. So it is not neces-
sary in practice with propellers of coarse pitch, to aim very closely
at some exact slip provided the propeller is so designed that under
conditions of service its slip is not too small. A propeller which
is too large, showing slip much below that for maximum efficiency,
will be very inefficient. On the other hand, a propeller may be
too small and work with slip a good deal greater than for maximum

efficiency without much loss of efficiency. It should be remem-

bered that the slips of Figs. 185 to 208 refer to propellers operating
in undisturbed water, and the apparent slip of propellers attached
to ships is usually less than the true slip.

When dealing with propellers of fine pitch ratio, say in the

neighborhood of unity, the question of efficiency as affected by
slip is complicated by the question of efficiency as affected by
pitch ratio. Thus in Fig. 190 we see that propeller No. 8, A cut,
of .25 mean width ratio and .8 pitch ratio has a maximum efficiency
of .632 at 15 per cent slip. From Fig. 194, propeller No. 13, A
cut, of .25 mean width ratio and i.o pitch ratio has a maximum
efficiency of .684 at 14 per cent slip and an efficiency of .632 at
about 31 per cent slip. In a given case, then, where we could fit a
propeller of the proportions of No. 8 working at maximum effi-

ciency, we could make an improvement if we could fit a propeller

of the proportions of No. 13 working below its maximum efficiency

provided did not exceed 30 per cent. This is a question of

its slip

very considerable practical importance. In the next section will

be given methods for determining the best combinations of pitch
ratio and slip for given conditions.

26. Practical Coefficients and Constants for Full-sized Pro-

pellers Derived from Model Experiments.

i. General Line to be Followed in Reducing Model Results. -

The results of the model experiments for the standard elliptical
3-bladed series will of course be of value in the case of any pro-
peller design. It should be carefully remembered, however, that

they cannot be applied blindly. We

have determined experi-
mentally the thrust and torque and deduced the efficiency of a
number of small propellers at a 5-knot speed of advance through-
out the range of slip likely to be found in practice. These small
propellers covered for 3-bladed elliptical propellers the range of
pitch ratio, mean width ratio, and blade thickness fraction likely
to be found in practice. We know that so long as cavitation does
not appear the Law of Comparison will apply satisfactorily and
that the results of the model experiments will apply to full-sized
propellers working under the same conditions as the models. But
in applying the results we must remember that they do not
hold for cavitating conditions, which will presently be considered

The models were tested in such a manner as to be practically

freefrom hull influence, and we know that for full-sized propellers
driving ships there are material mutual reactions between pro-
pellerand ship. The question arises whether we shall attempt to
take account of these reactions in reducing the model results or
consider them separately.
It is much better, and even simpler in the end, to attack the

problem in detail.
2. Reduction of Model Results. We have seen that by means
of a p8 diagram, as in Fig. 180, the experimental model results may
be reduced to a form convenient for practical applications. But
ifwe simply construct a p8 diagram for each model tested it will
be a very laborious process to locate and utilize the particular

diagram adapted to a particular case. So necessary to de-

it is

velop diagrams, by interpolation if necessary, such that the pri-

mary factors involved are readily determined. We have to deal
with efficiency, diameter, pitch ratio, mean width ratio and blade
thickness fraction.
These are too many variables to be covered directly on a single

diagram. The first three are the most important. Width and
blade thickness are not independent in practice. To do a given
work at given revolutions the narrow blade must be thicker than
the wide blade. So four p5 diagrams, Figs. 211 to 214, have been
constructed from the model results of Figs. 185 to 208. Figure 211
refers to blades having a mean width ratio of .20 and a blade
thickness fraction of .06. Similarly Figs. 212, 213 and 214 refer
respectively to mean width ratios of .25, .30 and .35 and blade
thickness fractions of .05, .04 and .03. We shall see later how to
make slight changes involved by other blade thickness fractions.
The application of the p8 diagrams is very simple :


where P is the power absorbed by the propeller of diameter d

feet at R revolutions per minute when advancing at a speed of VA
Then p is theprimary variable fixed by the conditions of the
problem. Contours of 5 are plotted above p for equal intervals

of pitch ratio and curves of efficiency for the same intervals.

When p is known we can determine very promptly for any
value of 8 the pitch ratio and efficiency. In addition to the con-
tours of d above p contours of slip are plotted in dotted lines.
3. MaximumEfficiency. The efficiency curves show many in-
teresting and significant features. For a short interval each pitch
ratio shows an efficiency greater than any other, and evidently if
our choice is free we should for a given value of p use the pitch
ratio corresponding to optimum efficiency. Hence, there is drawn
an enveloping curve maximum
efficiency touching the suc-
cessive efficiency lines for the various pitch ratios which has upon
it a scale of the pitch ratios for maximum efficiency.

In this connection attention may be called to the fact that the

portion of each efficiency curve which gives the best efficiency for
a given p is in general of an efficiency below the maximum efficiency
attainable with the pitch ratio. This is particularly noticeable for
the largest values of p. For all values of p above very small ones
it isbetter to use a propeller of relatively coarse pitch and work
it at a fairly high slip greater than that corresponding to its
maximum efficiency than to use a propeller of finer pitch and
work it at its maximum efficiency. This for the reason that for
propellers of pitch usual in practice decrease of pitch means fall-

ing off in efficiency.

The p8 diagrams bring out clearly some of the basic conditions

affecting propeller design.

Once we fix for a propeller the power, P, it is to absorb, its revo-
lutions per minute, R, and its speed of advance, VA, the value of p
is fixed. Now it is apparent from the diagrams that for a given
value of p there is a maximum efficiency beyond which we cannot

go. We may very easily fall short of it, but even if we adopt the

very best combination of diameter, pitch and blade area possible,

we cannot get beyond a limiting efficiency. The p8 diagrams of

Figs. 211 to 214 were deduced from experiments with models of

3-bladed propellers with elliptical blades having ogival sections.

Hence the limiting efficiencies shown in them are not exactly the
same types of propellers, though they are about as high
as for all

as for any known type. But there is no doubt that they indi-

cate well the general variation of efficiency with p for all types of

propellers in present use. While there is a maximum efficiency,

about p = 3, and the efficiency falls off on either side, the values of
p that are found in practice are almost never materially below 3, so
that in practice the larger the p the smaller the limiting efficiency.
It the high value of p produced, if we give low-speed vessels high

revolutions, that has hitherto prevented the application to cargo

vessels of turbines directly connected to the propeller. Thus, sup-
pose we had a destroyer propeller absorbing 5000 shaft horse-
power at 800 revolutions with a speed of advance of 30 knots.
For this case

The limiting efficiency for this value of p is about .65 which though
low is not impossible. If now we had a large single-screw cargo
and passenger which required 5000 shaft horse-power to
make 15 knots speed of advance and adhered to 800 revolutions
per minute the value of p would be

800 V cooo -

For this value of p the limiting efficiencywould be inadmissibly

low. To hold p at 11.5 the revolutions would have to be reduced
to 142 which would make an inefficient turbine. An alternative is
to hold revolutions at 800 and use multiple shafts. But in order to
make the p value for each propeller 11.5 only, it would be neces-

sary to divide the 5000 shaft horse-power between 32 shafts, which

is of course impossible.

Another fact of serious practical importance which the p8 dia-

grams bring out is that there is practically a lower limit to the
pitch ratio which can be used to advantage. At first the best
pitch ratio falls off rapidly with increase of p, but for large values
more and more slowly, and for no value
of p the pitch ratio falls off
of pwhich it would be advisable to use in practice is it desirable to
go below a pitch ratio of .9 or a little less.
The slip for the best all-round efficiency which is below .15 for
small values of p increases steadily, until it is seen that propellers

of a pitch ratio of .9 should be worked at over .30 slip. This is

real slip, not apparent slip.

It is interesting to note in this connection that the model ex-
periments indicate that the broader the blades the greater the
slip for the best results. Thus for a pitch ratio of i.o and the four
blade width ratios of .20, .25, .30 and .35 the best slips are respec-

tively .255, .265, .280 and .320. This is in accord with theoretical
4. Methods of Calculations. In order to facilitate the calcula-
tion of p in a given case there are given in Table XIV values of
VA .

be carefully borne in mind that VA is not the speed of

It should
the ship through the water but the speed of advance of the pro-

peller through the disturbed water in which it works. The differ-

ence between VA and V, the speed of the ship, will be considered
in connection with the wake factor.
The formula for 5 is

or, when 8 has been determined,

With a table of squares and cubes we can readily determine (PV A ) k

by taking the square root of the cube root of 'A', R* is simply PV

the square of the cube root of R. Hence the calculations re-
quired in connection with the use of the p8 diagrams are readily
5. Blade Thickness Correction. The
four p5 diagrams for the
standard series refer to a definite blade thickness fraction for each
mean width ratio. We have seen in Fig. 210 the effect upon the
efficiency of the standard series of variations of the blade thick-
ness. This effect isnot large enough to be of practical importance
in most cases. But variation of blade thickness will also neces-
sarily affect pitch ratio and diameter. Investigation shows, how-
ever, that the effect not large, and for blade width ratios from

.25 to .35, and for propellers of about the proportions for maxi-
mum efficiency, the average corrections required are shown in

Fig. 215. The curves of this figure give for various values of p the

percentages by which diameters and pitches determined from the

p8 diagrams must be modified when the standard blade thickness
fractions to which the pd diagrams correspond are departed from.
The corrections are small and in practice may often be ignored.
The standard p8 diagrams already take some account of thickness,
the widest blades being only half as thick as the narrowest, but of
course the actual blade thickness fraction in a given case is fixed

mainly by considerations of strength.

6. Four-bladed Propellers. The standard pd diagrams, Figs. 211
to 214, refer to three-bladed propellers. It would be desirable to
have similar diagrams from full experiments with four-bladed pro-
pellers, but lacking such they can be used with fair approximation
for four-bladed propellers. We have in Fig. 181 the relation be-
tween power absorbed, thrust and efficiency of three and four-
bladed propellers as deduced by analysis of experiments at the
model basin with propellers having quite thin blades of rather
broad tips. These may be taken as applying with reasonable
approximation to the elliptical blades.
Then the steps in a given case will be as follows:
1. Determine p in the ordinary way and then divide it by the
square root of the ratio between the coefficient A for a four-bladed
screw and for a three-bladed screw these ratios are given in

Fig. 181. Call the quotient P4.

2. Using p4 determine by the use of the proper p8 diagram the


proper diameter, pitch, etc., for a three-bladed propeller.

Then upon adding a fourth identical blade to the three-bladed
propeller we shall have a four-bladed propeller which will meet
the conditions.
For let P, R, and VA denote power to be absorbed, revolutions
to be made and speed of advance.
We have

P =

then P4 =

where r. is the ratio of the A coefficients from Fig. 181. Then

a three-bladed propeller based upon p4 will, at revolutions R and
speed of advance VA, absorb a power
But from Fig. 181 again

a four-bladed propeller identical as to diameter, pitch and blades

p =
will absorb r times the power of the three-bladed one, or X r P.

Hence the four-bladed propeller will absorb the power at revo- P

lutionsR and speed of advance VA- The relative efficiencies
may be obtained from Fig. 181.
Since once we know p, we can determine the relative diameters
of the three and four-bladed propellers; we can from each p8 dia-
gram for three-bladed propellers determine, by using Fig. 181 as
explained above, a figure giving ratios of diameter, pitch and
efficiency for three and four-bladed propellers. It is found, how-

ever, that as regards diameter and pitch the ratios are so nearly
the same for all widths that the results may be averaged in a simple
diagram (Fig. 216).
This gives curves of coefficients by which the diameter and
pitch of -a three-bladed propeller must be multiplied to determine
the diameter and pitch of a four-bladed propeller of the same type
of blades and mean width ratio that at the same revolutions and
speed of advance will absorb the same power.
Efficiency coefficients are also given. These are seen to be all
less than unity, indicating a loss of efficiency by adopting four-

bladed instead of three-bladed screws.

The pitch coefficient is less than unity throughout, so the
pitch of the four-bladed screw will be slightly less than that of
the three-bladed screw, but the diameter is reduced more than the

pitch, so that the pitch ratio of the four-bladed screw will be the
greater. The diameter coefficient in Fig. 216 should be regarded
as an upper limit. It will be feasible in practice to reduce the
diameter of the four-bladed screw four or five per cent more with-
out material loss of efficiency.

7. Two-bladed Propellers. It is evident that the methods above

may be utilized in order to apply the p8 diagrams for the three-
bladed propellers to two-bladed propellers.

In this case, however, the artificial value of p will be greater than

the original value.

Fig. 217 gives curves of coefficients, etc. It is seen that diame-

ter, pitch and efficiency are all increased. The gain in efficiency is

small, however, and there are practical objections to two-bladed

propellers, so that their use is seldom expedient. This point will
be discussed further in considering design of propellers.

27. Cavitation

i. Nature of Cavitation. The phenomenon known as cavita-

tion has been given muchattention of late years in connection with

quick-running turbine-driven propellers. It appears to have been

first identified upon the trials in 1894 of the torpedo boat destroyer

Daring which had reciprocating engines. When driven at full

power with the original screws this vessel showed very serious
vibration evidently due to some irregular screw action. The pro-

pulsive efficiencywas poor, the maximum speed obtained being

24 knots for 3700 I.H.P. and 384 revolutions per minute.
Mr. Sidney W. Barnaby, the engineer of the Thorneycrofts, who
built the Daring, came to the conclusion that at the high thrust

per square inch at which the screws were working the water was
unable to follow up the screw blades and that the bad perform-
ance of the screws was due to the formation of cavities in the
water forward of the screw, which cavities would probably be
filled with air and water vapor." So Mr. Barnaby gave the
phenomenon the name of cavitation. The screws which gave the
poor results had 6 feet 2 inches diameter. 8 feet 7! inches pitch and
8.9 square feet blade area. Various alternative screws were tried,
and the trouble was cured by the use of screws of 6 feet 2 inches
diameter, 8 feet n
inches pitch and 12.9 square feet blade area.
With these screws 24 knots was attained with 3050 I.H.P. and the
maximum speed rose from 24 knots to over 29 knots.
For the Daring cavitation appeared to begin when the screw
area was such that the thrust per square inch of projected area was
a little over u
pounds per square inch. For a time it was thought
that the thrust per square inch of projected area was a satisfactory
criterion in connection with cavitation and that the limiting

thrust per square inch of projected area found on the Daring was
generally applicable.
This, however, is not the case. Greater thrusts have been suc-
cessfully used and cavitation is liable to appear at much lower
thrusts. In one case within the author's experience cavitation
appeared when the thrust was about 5 pounds per square inch of
projected area, the tip speed being about 5000 feet per minute,
and in another when it was about 7.5 pounds, the tip speed being
about 650x3 feet per minute. There is little doubt that the prime
factors involved in cavitation are: the speed of the blade

through the water, which is conveniently measured by the tip

speed, and (2) the shape of the blade section.
2. Accepted Theory of Cavitation Inadequate. When we at-

tempt to explain just how or why vacuous cavities at the backs of

screw blades cause the serious loss of efficiency associated with cavi-
tation we encounter insuperable difficulties. Suppose, for instance,
the cavity is a vacuum and covers the whole blade back. Then the
thrust per square inch of projected area due to the vacuum on the
blade back would be between 14 and 15 pounds and the thrust due
to the face would be added to that. As cavitation will appear in
some cases at thrusts per square inch of projected area as low as 4

pounds, it is evident that in such cases there cannot be a vacuum

over the whole blade back and thrust in addition on the face.
But suppose the blade had a vacuum over a portion of the back

only. There would be no increase of thrust from additional suction

of that portion of the blade back, but neither would there be any
increase of torque due to that portion of the blade back. The only
loss of efficiency would be a small amount due to the propeller

working with a slightly higher slip, while the loss of efficiency

accompanying cavitation is very much greater than this.

Fig. 218 shows a propeller blade section advancing through the
water at an angle of slip of 3 degrees not an unusual angle.
There are three regions indicated:
1. The leading portion of the back, denoted by A.

2. The following portion of the back, denoted by B.

3. The face, denoted by C.
It does not appear possible that cavities can form at A. This

portion of the section contributes negative thrust, and although the

point of demarcation between the portion of the back contributing
negative thrust and the portion contributing positive thrust (suc-
tion) probably varies in position with speed through the water and
slip angle, it appears reasonably certain that A always contributes
negative thrust and quite probable that this negative thrust in-
creases indefinitely with the speed.
Qver B
a cavity will form when the speed is high enough. It
will probably be small at first, and as the speed is increased, cover
a greater and greater portion of the section back. It cannot cover
the whole back, however, because it cannot extend over A to the
leading edge.
As regards C it has been generally assumed that the thrust from
the face always increases with increase of speed of the section
through the water.
3. Possible Theories of Cavitation. Now how is it conceivable
that cavitation can cause a rather sudden loss of efficiency when
the section is pushed to a sufficiently high speed ?
A. It is possible that when a vacuum is formed at B this portion
of the blade contributes no more suction or thrust while the nega-
tive thrust at A
continues to increase with resulting loss of efficiency.
This explanation would seem to involve the further assumption that
by far the major portion of the thrust of a propeller is due to the
suction of the blade back.
B. It is possible that when a vacuum is formed at B it is spoiled
by obtained from the surrounding water and the suction of the

blade back is decreased. This explanation is possible only if, when

the water still hugs the blade back, it sweeps away any air which

is sucked out of the water, so that while the water is in contact with
the back it is possible for the latter to exert a suction approaching
that of a perfect vacuum. But when the water breaks away from
the back, air leaking into the space is carried away by entrainment

only from the rear of the cavity, where the water comes together
again; and when the rate of entrainment is equal to the rate of leak-
ing into the cavity there is a balance of pressure, and though there
is a partial vacuum in the cavity the pressure is much greater than

a complete vacuum.

C. It is possible that when cavitation sets in the thrust from the

blade face absolutely or relatively.
falls off

A, B and C above appear to cover the possible theories of the

phenomena associated with cavitation. Whether cavitation is due
to one or more of these explanations or to something different still,
can be satisfactorily determined by experiment only, either on
models or on full-sized propellers.
4. Experimental Investigation of Cavitation. Experiments
with cavitation using full-sized propellers have not hitherto been
made, except inadvertently. While no theory of cavitation should
be fully accepted until confirmed by full-sized experiments the ex-
pense of a general investigation with large propellers has been
hitherto prohibitive, to say nothing of the time required and the

practical difficulties in the way. Small scale or model experiments

on cavitation present special difficulties. For the law of com-
parison to apply in spite of cavitation it would be necessary to have
the pressure around the model in the ratio of the size to the pressure
around the full-sized propeller.

This requires the model to work in water whose surface is covered

by a partial vacuum, or in hot water which has a vapor pressure
partially neutralizing that of the air.
The Hon. C. A. Parsons has done some work using the latter
method, but little has been published of the results. There are
great practical difficulties in making experiments along this line,
except with very small models.
A second possible method of investigating cavitation experimen-
tally by means of models is to test the model, not at the corre-
sponding speed, but at the actual speed of advance of the full-sized

propeller. When done, the pressures per square inch at cor-

this is

responding points of propeller and model are the same, and if one
shows cavitation so will the other. This method is hardly prac-
ticable for the model of the propeller of a 33-knot destroyer, but
for propellers of slow and moderate-speed vessels experiments
could be made without serious difficulty or great expense, either
in a model basin or from a special testing platform in front
of a vessel. This method, however, has not been used in practice.
For model propellers of any size, say 15 inches to 18 inches in

diameter, it would require very powerful driving and measuring

A method is to use the propeller testing gear already in-
stalled in a model basin with small propellers of such abnormal
proportions and shape that they will show cavitation within the
limits of speed and revolutions available.
Some experiments along this line have been made at the United
States Model Basin.
obtain pronounced cavitation from small propellers 12 inches
to 1 6 inches in diameter, tested at speeds of advance not over 7
knots or so, it is necessary to make the pitch ratio much smaller
and the width of blade much larger than for
ratio of thickness to
the propellers used in practice. Sixteen-inch models representing

propellers of ordinary proportions will not cavitate satisfactorily

at low speeds of advance, and the experimental gear available was
not powerful enough to drive them at high speeds.
The results obtained with the fine pitch propellers appear, how-
ever, to throw some light upon the subject under consideration.
Figure 219 shows expanded blade outline and blade sections for a
1 6-inch model propeller of 6.4-inch pitch. Figure 220 shows curves
of thrust and torque for this propeller plotted upon slip for speeds
of advance of 5, 6 and 7 knots. The major portion of Fig. 220 is
from Fig. 10 of a paper by the author before the Society of Naval
Architects and Marine Engineers in 1904, but the curves for the

5-knot speed have been extended, and the curves for the propeller
reversed have been added from the results of subsequent experi-
ments. For the propeller reversed the nominal slip is figured from
the nominal pitch of the back as tested (the face before reversal).
Figure 220 shows conclusively that, so far as this propeller is con-
cerned, the thrust per square inch of projected area has little to do
with the cavitating point. At a nominal slip of 15 per cent there
is evidently cavitation at the 7-knot speed. At this point the thrust

is80 pounds, or almost 4.3 pounds per square inch of projected area.
At 5 knots, however, the thrust per square inch of projected area
at which cavitation begins is about 9 pounds.
Other conclusions might be drawn from Fig. 220, but more illu-
mination can be obtained from the results of trials of a small pro-

peller especially designed to show cavitation. This propeller was

14 inches in diameter and of 4.2 inches pitch. Its developed blade
outline and blade sections are shown in Fig. 221. At the points
A, B, C, D and E small holes were made on each blade connecting
to the shaft, which was hollow. The hole in the shaft communicated
in turn with a pipe forward of the hub, which led finally to a tank
under being a pressure gauge on the line and
air pressure, there

valves for turning on or cutting off the air pressure as desired.

When making trials one hole only was left open in each blade.
This apparatus measured suction or partial vacua with great facility
but had to be handled carefully to measure pressure. When mea-
suring suction, the air pressure was cut off, when the propeller itself
would quickly exhaust the air and the amount of vacuum was read
on the gauge. When measuring pressure, thevalve was barely

cracked, so that a small quantity of air was dribbling out all the
time through the hole where pressure was to be measured.
In this way the passages in the propeller were kept clear of water,
whose presence would have prevented obtaining the pressure at the
A gauge pressure of a pound and a half or so was sufficient to keep
the air passing out when the propeller was at rest or turning over
very slowly, and the difference between this initial pressure and the
gauge pressure shown while running was taken as pressure at the
In the early part of a run for pressure the air would stop coming
out of the propeller; it would accumulate in the pipe and the gauge
pressure rise until air again began to come out and the gauge became
steady. At the end of a run the instant the propeller began to slow
down the air would burst forth.
While the apparatus and methods described above for measuring
pressure and suction are certainly not of minute accuracy, they gave
consistent results which are believed to be reasonably accurate.
For looking at the propeller under the test there was fitted a
fixed disc with a small slot, and immediately behind it a revolving
disc with a similar slot, which was driven at the same speed as the

propeller. The propeller was illuminated by a searchlight and when

looking through the slot in the fixed disc the propeller was seen once

during each revolution always in the same position. The discs and
searchlight could be shifted so that either back or face of the pro-
peller could be observed.
Figure 222 shows for the propeller of Fig. 221 and three knots
speed of advance curves of thrust, torque and of pressure or suc-
tion at the points indicated. The curves are plotted upon nominal
slip and pressure and suction are measured in pounds per square
inch. A scale showing tip speed is also given.
Figure 223 gives the same data as Fig. 222 for five knots speed
of advance.
When watching the operation through the slotted discs any
cavities present were plainly visible and it was easy to trace the
development of cavitation.
At about 3000 feet tip speed cavities appeared at the following
portions of the back and the leading portions of the face. The
cavities appeared first on the face, as might be expected from Figs.
222 and 223, which show that the suction at A is always more
intense than at D.
The cavities first show themselves near the blade tips and creep
in toward the center as speed is increased.
In Figs. 222 and 223 the thrust has returned to zero, when the
tip speed is between 5000 and 6000 feet per minute. When this
is the case the cavities at the back of the blade extend in from the

tip about two-thirds of the blade length and near the tip cover
nearly two-thirds of the blade back.
On the face under the same conditions the cavities extend along
the leading edge practically in to the hub and near the tip from the
leading edge to the following edge.
5. Theory and Cause of Cavitation. From the experimental
curves of Figs. 222 and 223 and observation of the cavities it is
obvious that the cavities at the rear of the blade do no harm. It
is the cavities on the driving face which grow rapidly as tip speed
is combined with the negative thrust of the leading por-
tions of the blade back that stop the increase of thrust and then

actually cause it to decrease to zero and below.

These conclusions apply strictly to the 14-inch model propeller
of somewhat abnormal type shown in Fig. 221, but it seems reason- .

ably certain that they apply more generally, and that harmful cavi-
tation is due not to cavities at the backs of propeller blades, but to
cavities at their driving faces.
When we seek a cause for these cavities, it seems fairly obvious.
Fig. 218 shows the section of a propeller blade advancing with a
slip angle of 3 degrees, which is not an exceptionally small angle,
as is evident from Fig. 170. But the face C, advancing through
the water at an angle of 3 degrees, is associated with the leading

portions of the back, whose direction is such that it is advancing

through the water at an angle of over 20 degrees. Fig. 63 shows
diagrammatically the nature of the motion of water past a plane
with a sharp edge. In the case of the propeller we have virtually
two planes in associationnamely, the face and the leading portions

of the back. Considering the face alone, the water tends to cascade
around the leading edge from front to back. Considering the back
alone, the water tends to cascade around the leading edge from
back to front. Actin? In association, the back of the blade with
an inclination of ;^er 20 degrees overpowers the face with an incli-
nation of 3' degrees, and as a result the water cascades from the back
of the blade to the face around the leading edge, causing first eddies
and then cavities on the face of the blade.
In regarding the leading portion of the propeller blade as made
up of two planes, we should remember that the motion at each point
is circular, not linear. A plane in linear motion can drag a good
deal of dead water behind and water brought to rest relatively

to the plane passes aft again without any motion across the plane.
The propeller blade is moving in a circle and cannot carry water
" "
with it in the shape of dead water for any distance. Centrifu-
gal action would rapidly throw it out, and no doubt strong centrif-
ugal force acts upon the water which is brought nearly or entirely
to rest relatively to the blade by impinging upon the leading
It possible that this strongly localized centrifugal force plays

a part in causing cavitation.

It is evidently necessary to consider separately the cavitation
which appears over the backs of propeller blades and the cavitation
which appears over the faces.
The former is not seriously objectionable. If the cavities at the

blade backs were perfect vacua they would be helpful rather than
harmful. It is seen from Figs. 222 and 223 that for model propel-
lers in the fresh water of the model basin these cavities do approach
perfect vacua. Sea water contains a good deal of occluded air, and
it may be that for full-sized propellers in sea water the cavities are

more or less filled with air. But, even so, the air could be pumped
out without serious difficulty. Hence we may conclude that cavities
at the rear of a blade are not an insuperable bar to efficiency. This
is fortunate, for there is no question that when a curved surface,

such as the back of a propeller blade, is driven through the water

at a sufficiently high speed, cavities are necessarily formed over its

rear portions.
The case of the cavities over the blade faces is different. These
have no redeeming feature. In the they are due to an
first place,

edge angle so large as to produce large negative thrust from the

leading portion of the back of the blade. In the second place, they
nullify the thrust which the blade" face would otherwise contribute,
and, things considered, are obviously fatal to efficiency.

Hence, it is essential to efficiency to minimize or avoid entirely

face cavitation. The method which has been most used with satis-
faction in practice consists in fitting very broad blades so that the
thrust per square inch of projected area is kept below a limit found
to be safe by experience. But the thrust per square inch of pro-

jected area is not the primary feature causing cavitation. Tip

speed and blade section are without doubt the main factors. Still,
for a given type of propeller the thrust a function of tip speed

and blade section, and hence might be used as a gauge of cavitating

conditions. Thus Barnaby, for the type of propeller used on the
Daring, found that with a tip immersion of one foot, cavitation
showed up when the thrust per square inch of projected area was
above 1 1 pounds. The trouble with this method is that the limit-
ing thrust permissible would have to be determined for each type
of propeller.
6. Reduction of Cavitation by Broad Blades. From the theory
of cavitation set forth above the advantages of a wide, thin blade
are obvious. It has a smaller edge angle, so that it can be driven

to a much higher tip speed than a narrow blade without causing


face cavitation. Also after face cavitation begins it spreads slowly

with increase of tip speed so that the wider the blade the greater
the area of the face whose thrust is not nullified by cavitation.
In fact, if the blade is so wide that the manner of the water leav-
ing not materially modified by cavitation, the thrust will not
it is

be materially modified even if there is a cavity over the leading

portion of the face. This result is readily explicable. Thus, sup-
pose we have a cavity at the leading portion of a blade face. The
vacuum results in the water being impelled toward the face, forward
momentum being communicated to it. If the face is sufficiently

wide, the water will impinge upon it again. Through the loss of
its momentum it will communicate a corresponding thrust to the

blade, and then will pass from the blade, if it is wide enough, in
nearly the same manner as if there were no cavitation over the
forward portion of the face. Hence, the net change of velocity and
resulting thrust will not be much affected by the cavitation. But
if the blade is so narrow that the face cavity extends nearly to the

following edge there will not be enough blade beyond the cavity to
absorb the forward momentum of the water and direct it again in
the way it should go. With the wide blade the loss of pressure
on the leading portion of the face is nearly made
due to cavitation
up by additional pressure on the following portion of the face.
With the narrow blade there is virtually no following portion.
Figures 224 and 225 show experimental results which indicate the
advantages of breadth of blade in preventing harmful effects from
cavitation. Two model propellers of the same pitch ratio
1 6-inch

0.4 and blade thickness fraction, but of mean width ratios of

.125 and .275, were tested with smooth backs and with strips secured
to the backs, as indicated in the figures. The sections shown were
taken in each case at two-thirds the radius. The curves in each
case refer to a 5-knot speed of advance. Neither propeller showed
harmful effects of cavitation with a smooth back. With the strip
attached the narrow-bladed propeller showed pronounced cavitation,
while the broad-bladed propeller showed none, though its strip was
materially larger than that of the narrow-bladed propeller. As
might be expected, the torque is much increased by the presence of
the strip. But until cavitation appears the thrust of the narrow-

bladed propeller is but little reduced by the strip, and for the broad-
bladed propeller the thrust actually increased by the presence of

the strip. Upon the theory of cavitation which has been set forth
a reasonable explanation of the peculiar features of Figs. 224 and
225 is as follows:
The strips increase the negative thrust on the leading portion
of the blade back in each case, increase the suction or cavitation
of the following portion of the blade back, thus increasing thrust,
and cause face cavitation over the leading portion of the blade face.
The net two former actions is small or even results in
result of the
an increased thrust. But when face cavitation is set up strongly,
the narrow blade breaks down, while the broad blade holds its own,
because the face cavitation over the leading portion of the face is

neutralized by the action of the following portion of the face.

7. Cure for Cavitation. We have seen that the wide blade of
usual type has two advantages from the point of view of cavitation.
Its smaller edge angle will allow high tip speeds to be reached with-
out cavitation, and when cavities do appear the tip speed can be
still further increased without the harmful effects due to the face

cavities, which are usually characterized by the term "cavitation.'*

Now we do not mind cavities on the back of the blade, so the ques-
tion whether it is possible fully to cure harmful cavitation depends

entirely upon whether it ispossible to avoid entirely face cavitation.

The difficulties in the way of this are practical difficulties of con-

struction. Thus, if we could make propeller blades without thickness,

there would be no face cavitation. The water would cascade around
the leading edge from front to back. There would be back cavita-
tion only, and solid water over the face. But we cannot make pro-
peller blades of no thickness. The best we can do in practice is to

approximate to the ideal plane along the leading edge, making the
face straight, or very slightly convex, and the leading portions of
the back hollow, as indicated in Fig. 226, and keeping the edge
angle down as close as possible to the slip angle.
It might seem that the edge angle could be made double the slip
angle without danger of face cavitation, since when so made the edge
would part the water evenly. But the slip angle is an average angle,
and usually at some part of its revolution the blade of an actual

propeller will have a slip angle but little if any greater than half
the average value. Another reason for making the edge angle as
small as practicable is the fact that no matter how sharp the edge
ismade it is not a mathematical edge, and when advancing at enor-
mous speed through the water will show slight cavitation if it is
attempted to split the water evenly on each side. Hence, the en-
deavor should be to have the water naturally tend to cascade around
the edge from face to back.
It might seem that this could be accomplished without extreme

sharpening of the leading edge by making the leading portion of the

face convex, as indicated in Fig. 227. This is true, and a propeller
so shaped would not show face cavitation near the leading edge,
but with even a moderate convexity of the face it would show
severe cavitation over the following portion of the face. There
was a case of a United States battleship whose propeller did not
differ materially in dimensions, etc., from those of her sister vessels,

but had sections which were abnormally curved at the leading por-
tion of the face, as indicated in Fig. 228.
This vessel showed over a knot less speed than her sister vessels
for the same power, and although her tip speed was only about
6000 showed very
feet per minute, there is little question that she
serious face cavitation. not
It ispossible to say what convexity is
permissible in a given case without cavitation, but it is certain that
the higher the tip speed the smaller the permissible convexity, and
for tip speeds of 10,000 feet and over it probably should be very
small indeed. Pending careful full-scale experiments on this point,
the safest plan is to avoid convex blade faces for propellers of high
tip speed.
need hardly be said that it is not easy to make hollow-backed

propellers with leading edges as sharp as a knife. It is advisable

to use cylindrical ribs on the back, extending from the leading edge
to the thicker portion of the blade. If the leading edge is serrated

with a rib extending to the point of each tooth, the blade edge need
not be quite so sharp. Such a form of edge seems to get through
the water with less tendency to face cavitaiion, and when this does
set in it seems to confine itself to rather narrow rings, starting from
the angles where the roots of the serrations join.

The ribs on the back must be well sharpened where

of course

they cut the water. They increase back cavitation, but that is not
a very serious matter.
While the prevention of face cavitation is essentially a question
of the extreme leading portion of the blade back, the blade should
not thicken so rapidly as we pass aft from the hollow portion that
owing to its angle of action there is large negative thrust.
This is of course always objectionable, but particularly so when
there pronounced back cavitation. After this has set up, the

suction of the back does not grow so rapidly as before with increase
of speed, and hence negative thrust, which continues to increase

indefinitely with speed, should be avoided with peculiar care.

The practical conclusion in this connection is that blades made
hollow-backed to avoid cavitation should not be of narrow type
but fairly wide say from .30 to .35 mean width ratio in
order that they may be made fairly thin in the center.
Such blades should avoid cavitation without the excessive widths
which are necessary with blades of ogival section and which involve
material loss of efficiency through large blade friction.
8. Pressure Due to Blade Edge Speed. In connection with the
question of cavitation it is interesting to note that at the tip veloci-
ties ofmodern high speed propellers enormous pressures are liable
to be set up upon the leading blade edges. Suppose we have a
small plane advancing through water perpendicular to itself. The
maximum pressure upon it is that due to a head equivalent to the
velocity, the formula being


where p pressure in pounds per square foot, v is velocity of advance


in feet per second, w is weight of a cubic foot of water and g is the

acceleration due to gravity. If we assume that at a blade edge

there is always a small portion which is virtually a plane surface,

it follows that the motion of the blade through the water will cause

at its edge the pressure given by the above formula.

Table XV
shows for various blade edge velocities in feet per
minute, g being taken as 32.16, the corresponding pressures in salt

water weighing 64 pounds to the cubic foot. The pressures are

expressed in pounds per square inch.
When we consider in Table XV
the very rapid growth of blade
edge pressures with velocity and the very high pressures reached
when 10,000 feet per minute and over, it is obvious
the velocity is

that for high-speed propellers the area of blade edge over which
such pressures are set up must be reduced to a minimum. In former
days propeller blades were often made of elliptical section, and even
now, for fairly high-speed propellers ogival blades are frequently
finished with a quarter round. Such blades will certainly break
down by and quick running propellers
cavitation at high-speeds
should by .means have sharp leading edges. It is difficult to make

an edge which is mathematically a sharp edge, but the more nearly

this is approached the better.

28. Wake Factor, Thrust Deduction, and Propeller Suction

Hitherto the ship and the propeller have been considered apart.
It is necessary now to take up their very important reactions upon
one another when the ship is being driven by its propeller or pro-
i. of Wake.
Components Owing to its frictional drag upon the
surrounding water there is found aft in the vicinity of the ship a

following current or wake, called the frictional wake, which is in most

cases greatest at the surface and in the central longitudinal plane
of the ship and decreases downward and outward on each side.

Superposed upon the frictional wake there is a stream line wake,

caused by the forward velocity of the water closing in around the
stern. This also will be greatest at the surface and center and
decrease downward and outward, though its law of decrease will be

different from that of the frict onal wake.

Superposed upon the two wakes above we have the wave wake.
If there is a wave crest under the stern, the water is moving forward
with velocity which decreases downward from the surface and, prob-
ably in practical cases, decreases slightly outward from the center.
Under a wave hollow the velocity is sternward, the wave wake

velocity in this case may be said to be negative, the wake being

regarded as positive when its velocity is forward.


There is a final factor, often ignored, which will be considered in

more detail later in connection with shaft obliquity. The water
aft not flowing exactly parallel to the shaft. It rises up behind

the stern and closes in horizontally, thus causing the slip of a pro-

peller blade tobe greater than the average over one portion of its
revolution and less than the average over another. This condition
of affairs does not materially affect the wake action, except in certain
cases that will be considered later. For the present we will consider
the wake proper made up of the three components enumerated
2. Effects of Wake. The propeller of an actual ship does not
work undisturbed water, but in water which has a very confused

motion. The wake velocity varies over the propeller disc at a given
speed, and at a given point of the disc varies with the speed. It
is necessary to assume a uniform velocity of wake over the screw
disc. This velocity of wake may conveniently be expressed as a frac-
tion of the velocity of the ship, the ratio being called the "wake
fraction and denoted by w. The wake was first explored by R. E.
Froude, who published some methods and results as long ago as 1883
in a paper before the Institution of Naval Architects. Froude used
model propellers behind ships' models. Suppose the speed of the
ship model is V. If the model screw is tested at given revolutions

separate from the model at a speed of advance V into still water,

we get a certain thrust and torque.
Suppose, now, keeping the revolutions constant, the model screw
is tested behind the ship model. The thrust and torque are changed
and are the same as would be found at the constant revolutions at
a speed of advance Vi, say, into still water. V\ is nearly always
less than V. So the wake behind the model at the speed V is equiv-
alent, so far as thescrew is concerned, to a uniform following cur-
rent of velocity V V\ or wV. The thrust and torque of the
screw are then those appropriate to a speed of advance of V\. The
power absorbed is the same as if the screw were working in undis-
turbed water with speed of advance V\. But if T denotes the
thrust, the useful work as far as the ship is concerned is not TV\
but T V. Hence the efficiency or ratio between the useful work and

power absorbed is, if V is greater than FI, greater than in undis-


turbed water, the ratio being -
The fact is that the following
wake assists in pushing the ship ahead, using the propeller as the
Thrust Deduction and Hull Efficiency.
3. While the ship acts
upon the screw through its wake, the screw acts upon the ship
through its suction.

Through its suction, the resistance of the ship is virtually in-

creased beyond what it is without the screw. This is a cause of
increase of power absorbed in propulsion. If R is the resistance

of the ship at speed V, and T the screw thrust required to drive

the ship at speed V, we have T greater than R. The quantity
T - R is called the thrust deduction, being the difference be-
tween the actual thrust and the net thrust or tow-rope resistance.
It is usually denoted by tT, so that R = T(i t) and i / is called

the thrust deduction factor, t

being called the thrust deduction

Suppose, now, we have a propeller absorbing a certain power, P,

at certain revolutions per minute and driving a ship at speed V.
In undisturbed water the propeller when absorbing the same power
at the same revolutions would have a speed of advance Vi, and its
would be a definite quantity, e say. Its thrust is T. De-
note the effective horse-power necessary to propel the ship by E
and its resistance by R. Then E is not equal to eP, as it would
be if there were no wake or thrust deduction, but to eP X X
T V\
The expression X is called the hull efficiency, and its two factors
T V\
and are called respectively the thrust deduction factor and
T V\
the wake factor. Since R = T (i /) and Fi= V (i w) we have

the hull efficiency = ~X rr = ~

T Vi i w
Froude expressed the wake as a fraction of Vi, the speed of ad-'
vance, not V, the speed of the ship. Calling this wp Froude denoted ,

the wake factor by i -f- wp where wp is the "wake percentage."
" "
There are some advantages in using the wake fraction as already
denned, but care must be exercised not to confuse it with Froude's
"wake percentage." The relation connecting them is

In most cases the hull efficiency does not depart greatly from
unity, the thrust deduction factor i t
being less than unity, and

the wake factor than unity.

i w greater
This readily understood when we reflect that the more favorably

a screw is situated to catch the wake the more direct its suction as
a rule upon the after part of the ship. Single screws, for example,
may be expected to show larger thrust deductions and wake factors
than twin screws. Also the stream line wake is increased by full

lines aft, but the part the stronger the propeller

fuller the after

suction upon it and the larger the thrust deduction factor.

4. Variations of Wake Fraction and Thrust Deduction. The
wake fraction and thrust deduction are affected by many considera-
tions, and in the present state of our
knowledge the actual values
in a given case can seldom be estimated accurately without special
model experiments.
The most comprehensive information in this connection available
at present is contained in a paper read at the 1910 Spring Meeting
of the Institution of Naval Architects by W. J. Luke, Esq. This
paper contains data as to the wakes and thrust deductions of models
of various vessels that had been previously published, mainly by
Mr. R. E. Froude, and gives a great deal of valuable new infor-
mation obtained at the John Brown and Company's experimental
tank at Clydebank, Scotland. These experiments were made with
a single model 204 inches long, 30 inches broad, of 9 inches mean
draught, displacement 1296 pounds in fresh water and having .65
block coefficient. All variations of propellers, etc., were tried with
the bare hull and many with propeller bosses or brackets inclined
225 degrees from the horizontal. In addition some special experi-
ments were made with bosses at other angles, ranging from horizontal
to vertical.

In what may be termed the standard conditions, two three-bladed

model propellers 6 inches in diameter, of 7.2-inch pitch with straight
elliptical blades were used with centers i inches forward of the
after perpendicular and 5 inches from the center line.

Experiments were made varying separately speed of vessel, pitch

ratio and diameter of propellers, fore and aft and transverse position
of propellers, number and area of blades, etc.

Briefly summarizing the main results of the twin screw experi-

ments, which were always made with both outward and inward
turning screws, Luke found that variation of number and area of
blades had no appreciable effect upon wake factor and thrust

Change of pitch ratio produced changes of secondary importance

for the bare hull, both wake and thrust deduction increasing slightly.
With the 22\ degrees bossing the changes were slight and much as
before with outward turning screws, but with inward turning screws
the wake fell off with increase of pitch.
Changes of diameter caused material changes in wake and thrust

deduction, but Luke concluded that they were due as much to

changes in clearance between hull and propeller as to the changes
in diameter per se.

Change of speed of vessel resulted in practically no change in

thrust deduction, but whether with bare' hull or bossing the wake
fell off steadily with increase of speed, the wake fraction decreasing

with the bare hull and propellers in standard location from about
.19 for speed-length ratio of .6 to .1452 for speed-length ratio of i.o.
In the paper the wake is characterized by the wake percentage
values following Froude. These have been converted to wake frac-
tions as already defined. For a speed-length ratio of .8, about what
such a vessel would usually be driven at in service, the wake fraction
was .167 for inturning screws and .173 for outturning screws, the
thrust deduction being about .155 in each case.

With the bossing the thrust deduction was still practically the
same with out- and inturning screws and varied little from .16.
The wake fraction fell off with the speed as with the bare hull, but
the wake was materially greater for outturning than for inturn-

ing screws. For the .8 speed length ratio it was .191 instead of

.173 for outturning screws and .146 instead of .167 for inturning
Luke's experiments show clearly that for the model tried the most
important factor affecting wake and thrust deduction is the location
of the propeller with reference to the hull. Thus with the bare hull
and the 6-inch propeller the results were as follows :

Center of propeller from center of model


- -
broadly speaking, wake factor and thrust deduction factor
i w
i t were reciprocals or t. w =
The data given by Luke confirms
this, and shows we may, so far as present knowledge
also that

goes, reasonably assume wake fraction to vary linearly with block

Then from the data published by Luke we may say with reason-
able approximation iv = .2 .55 b
= I, where w is wake frac-
tion, / is thrust deduction coefficient and b is block coefficient.
This formula ignores the matter of screw location, but may be taken
as applying to screws about abreast the after perpendicular and
with centers about 1.2 the radius from the center line.

For lesser clearance w will be greater and t will also increase

somewhat, but the formula is and can be, from the available data,
only a rough approximation.
For center screws in the usual position the approximate formula
indicated is" w = .05 + -5 b.

Data is not available for a formula for t for center screws, but
Luke's experiments would appear to indicate that for them t would
be increased but little over its value for twin screws. It follows
that if the hull efficiency is unity for twin screws it is somewhat
over unity for single screws, particularly for full vessels.
The formulae above apply to the bare hull or to vessels fitted with
struts or bossing which does not interfere with the natural water
It should be remembered that they are deduced from model
experiments and will nearly always exaggerate the wake of the full-
sized ship. It is desirable, however, if we cannot determine the
wake accurately, to overestimate it rather than underestimate it.
If it is overestimated, the engines on trial will turn somewhat faster
than estimated, which is generally allowable. If it is underesti-
mated, it may be impossible to run the engines up to the designed
speed without decreasing propeller pitch or reducing propeller
6.Approximate Determination of Wake Fraction. Since the
wake is explored by trial of model screws working behind models of
ships the question naturally arises whether we cannot gain some

lightupon the subject from trials of full-sized ships. Analysis soon

makes it evident that the apparent slip of propellers on trial is often
very much below what must have been the real slip. We know
that in any case the power absorbed by a given propeller advancing

through undisturbed water depends only upon the revolutions and

the speed of advance. For an actual propeller advancing through
the water disturbed by the ship we can reasonably reduce the actual
disturbance to an equivalent uniform motion. Throughout the
range where the Law of Comparison holds we can determine for
any propeller for which we have model experiments the relations
between power absorbed, revolutions and speed of advance. Hence,
if we know any two of these quantities, we can determine the third.

Now from the results of trial of a vessel we know corresponding

values of indicated or shaft horse-power, revolutions and speed of
vessel. The shaft horse-power is practically the power, P, absorbed
by the propeller, and from the indicated horse-power P can be
estimated with reasonable accuracy. Hence, although we do not
measure VA directly, we can estimate it from the power and revolu-
tions if we have model experiments with the propeller and
the Law of Comparisonholds, and knowing VA and V we can deter-
mine the wake fraction. The reduction of the results of model
experiment to a form convenient for this application is simple. We

have seen that we may write P = A - A

where P is power absorbed

by the screw, d is diameter in feet and V A is speed of advance in

knots. A is a coefficient independent of size and speed and de-
pending only upon the slip and the proportions and shape of the


So let us write P = S{/ PR \

where p is pitch in feet and R
Viooo/ p
Then S =

denotes revolutions per minute. I

) jj6
P and is like
\ pR I a
A, a coefficient independent of size and speed and depending only
on the slip and the proportions, etc., of the propeller.
From models we can readily determine
experimental results with
a curve of 5 plotted on the slip. Thus, for a 1 6-inch model with a
speed of advance of 5 knots we have 5 =.3129^ (i s) where Q

is torque in pound-feet. Or we may determine 5 from a curve

101.33 1000 (ioi.33) 3 \ioooj p

Whence S =.9610% A (i s)

Fig. 229 shows curves of S plotted on s for the four propellers of

Fig. 179. Now suppose we have a full-sized propeller similar to

the model of .0448 blade thickness fraction and 18 feet in diameter,

making 120 revolutions per minute and absorbing 12000 horse-
power. Its pitch will be 21.6 feet. Then from the data of the
/ 1 OOO\ &
full-sized screw S= ( ) ^P = 6 2.552. From Fig. 229, for the
\ pR I a
propeller in question, when S 2.552, s
= .2340. So the true slip
of this propeller would be .2340, and its true speed of advance,

VA = iQ-593- Suppose the speed of the ship V is

21 so that the apparent slip, 5', is .1790.

Then V = ^R ^ ~^ = 21 knots.

The wake =V VA = -
- = 1.407 knots.


ITT- i r 4.-
fraction = V VA- = I0l
= -S s
= .0670.


very easy to derive curves of

It is S from the Standard Series
results of Figs. 185 to 208.

Figures 230 to 233 show contours of slip plotted on S and pitch

ratio for four blade widths and the blade thickness fractions indi-
cated. For propellers closely resembling the Standard Series these
figures may be used in connection with accurate trial data to obtain
a reasonable approximation to the wake so long as there is no cavi-
tation. The propeller power, P, however, must for reciprocating
engines be estimated from the I.H.P. Methods for this will be con-
sidered under analysis of trials.

These figures be used, however, to obtain rough approxima-

tions to the wake for propellers very different from the Standard
For three-bladed propellers with oval blades and extra wide tips
the correct values of 5 will be somewhat less than in the figures, but
the difference for practical propellers will not be great. In order
to use Figs. 230 to 233 for four-bladed propellers we need only
divide the actual propeller power, P, by the proper power ratio for
four blades, obtained from Fig. 181. We thus obtain approximately
the power absorbed by a three-bladed propeller having blades iden-
tical with the four-bladed propeller and working with the same

revolutions and speed of advance.

From this we determine 5 and
use Figs. 230 to 233 as before.
It will be found in practice that the methods above for estimating
the wake from full-sized trials will generally give values that seem
too low. We know that the wake values for a full-sized ship should
be less model, but another factor present at times and
than for its

tending to lower the wake deduced from the S value is a slight

failure of the Law of Comparison connecting model and full-sized

propeller. We know that the Law of Comparison fails when a pro-

peller breaks down by cavitation, but it is probable, particularly
with blunt-edged blades, that there is more often than might be
supposed a certain amount of eddying in the operation of the full-
sized propeller not found in the operation of the model. This might
not seriously reduce efficiency and would manifest itself mainly by
a slip of the full-sized propeller somewhat larger than would be
inferred from the model results. The wake deduced from the S
values would be correspondingly reduced.
The 5 value method should not be used when the wake can be
investigated by model experiments. Lacking model experiments,
we can roughly approximate to the wake by the formulae already
There great need for a systematic and thorough experimental

investigation of the question of wake, following the lines of Luke's

experiment, which will enable it to be closely estimated in any prac-

tical case likely to arise. But there is a mass of accummulated

trial data extant for vessels whose models never have been, and

probably never will be, tested, and it is worth while for those possess-

ing it to investigate the wake by a method which is

fraction even

only roughly approximate. For practical purposes the wake frac-

tion of a vessel seldom requires to be determined with minute accu-

racy. It is principally of use for settling the diameter and pitch

of the screw, and neither these nor the efficiency will often be much
affected by a moderate error in the wake fraction.
If by use of Figs. 230 to 233 we find a certain wake for a vessel
of a given type, we can use this for a vessel of the same type with
similar propeller location, and for the purpose of determining diam-
eter and pitch of screw it will make little difference whether the
nominal wake from Figs. 230 to 233 is the real wake or departs
materially from it. Whatever the departure, it will be practically
the same in the two cases.
7. Effect of Shaft Brackets upon Wake. Reference has already
been made to the apparent effect upon the wake of the direction
of flow of the water aft.
This has a marked effect when large shaft brackets are fitted
which modify the natural flow of the water.

Thus, if a shaft bracket is fitted with a wide horizontal web, it

interferes seriously with upward flow aft and the water closes in
with a much stronger horizontal motion or current inwards than
otherwise. The conditions over the lower half of the propeller disc
are somewhat, but not very seriously, modified from bare hull con-
ditions, much greater modifications occurring over the upper half
of the disc. Considering the upper blades, the effect of the inward
flow of the water is materially to increase the slip angle for outward

turning propellers where the upper blades are moving against the
current, while for inward turning screws with the upper blades mov-
ing in the same direction as the current the slip angle would be
decreased. Hence, we may expect a large horizontal shaft bracket
materially to increase the apparent wake for outward turning screws
and to decrease it for inward turning screws.
A case in point that of the Niagara II, a steam yacht 247' 6"

X 36' X i6'4^" draught and 2000 tons displacement.

This vessel had a Lundborg stern, involving wide horizontal
shaft brackets, and her deadwood aft was not cut up.

She had two six-hour trials under similar conditions, except that
the screws were interchanged, being inward turning on the first trial
and outward turning on the second. While the horse-power was
not accurately determined, it was closely estimated at 2100 with
inward turning screws and 1950 with outward turning screws.
Nevertheless, with inward turning screws the average speed was
12.8 knots with an apparent slip of 26.4 per cent, while with out-
ward turning screws the average speed was 14.12 knots with an
apparent slip of but 13.3 per cent.
This marked difference in apparent slip can be due only to the
fact that the horizontal shaft webs force a strong inward motion
of the water above them along horizontal lines, and while this motion
is not a wake, being transverse or perpendicular to the line of

advance of the ship, its effect upon the upper blades of the pro-
peller is equivalent to a positive wake for outturning screws and a
negative wake for inturning screws.
It would seem that the lower blades are not much affected, such
action as there may be upon them being much less than that upon
the upper blades.
Luke's paper already referred to, gives most interesting and in-
structive results of a model investigation of shaft bracket angles
and direction of screw rotation. The model was the same as already
described, 204 inches long, 30 inches broad, 9 inches draught, 1296
pounds displacement in fresh water, having a block coefficient of .65.
The model screws were three-bladed, 6 inches in diameter, 7.2 inches
pitch, having straight elliptical blades. Their centers were 5 inches
out from the center line of the model and ij inches forward of the
A.P. Brackets were fitted at angles ranging from horizontal to
vertical and the model tested with inturning and outturning screws,
the screws and their positions remaining unchanged as the shaft
bracket angles were varied. The results are summarized on the
following page.
These results show relatively wake with
enormous variations of
variation of bracket angle and direction of turning and make it
clear that under some conditions the virtual wake due to obliquity
of water motion may overshadow the real wake or forward motion.
It is obvious that for a given real wake outturning and inturning

screws should give practically the same derived wake. We see, how-
ever, that with horizontal brackets the wake fraction is about 2\
times as great with outturning screws as with inturning screws while
with vertical brackets the wake fraction with inturning screws is
nearly four times as great as with outturning screws.

suction of 1 6-inch model propellers was measured over the surface

of a vertical plane, parallel to the propeller axis, which could be set
at various distances from the propellers. Figure 234, showing the
variation of pressure along various horizontal lines of the plane when
set f inch from the tips of a propeller of 1 6-inch diameter and 16-
inch pitch which was working at a nominal slip of 30 per cent, is

typical of all the results.

Now a necessary result of this suction is that it draws the water
inward toward the propeller axis and aft toward the disc. An impor-
tant fact, which seems to have been generally ignored, should be

pointed out. When

a propeller works with sternward slip velocity
of the water, the supply of water necessary to allow slip velocity
comes ultimately from the free surface. For referring to Fig. 235,
which indicates a submerged propeller, consider an imaginary plane
X Y perpendicular to the shaft axis and just forward of the screw
disc, as indicated by the dotted line. But for the screw action all
the water in that plane would be at rest. Owing to the screw action
the water is flowing aft through the screw disc and forward is
flowing from all directions toward the disc. Now the water flowing
through the plane does not leave a vacuum behind it; and particles
of waler flowing toward the disc from points forward of the plane
cannot leave vacua behind them. Their places must be taken by
other particles of water. Where can these particles come from ?
The water being practically incompressible, there are only two pos-
sible sources of supply. It is possible to conceive that the water

flowing aft through the plane just forward of the screw disc spreads
out astern, and finally to an equal amount flows forward again
through the plane. In other words, the suction draws a certain
amount of water through the plane and the thrust behind the pro-
peller forces an equal amount across the plane in the opposite direc-
tion at points some distance from the disc. This action goes on
when a screw is operated with no speed of advance as in dock trials.

Careful study of the action of advancing screws, however, indicates

clearly that in this case the water to take the place of that sucked
to the piopeller disc simply flows downward from the surface, pro-

ducing a depression of the surface, which advances with the speed

of the propeller. Figures 235, 236 and 237 show results of an experi-

mental investigation of this question made at the United States

Model Basin. Two 1 6-inch propellers of identical blade profiles, as
indicated, one with 12. 8-inch nominal pitch, the other with 19.2-
inch nominal pitch, were operated as indicated with their tips
8 inches below the surface and the resulting surface depressions for
5-knots speed of advance and various slips observed. It is seen
that contour lines in the depression over the propeller are approxi-
mately circular. The point of maximum depression is in each case
a little astern of the propeller, and as to be expected, the greater

the slip the greater the depression also the finer the pitch the greater

the depression. This too is to be expected. The propeller of fine

pitch exerts much the greater thrust for a given slip and speed of
advance. Hence the actual sternward velocity communicated to the
water is greater for the propeller of fine pitch than for the propeller
of coarse pitch, and the surface depression greater accordingly.
As a result of the fact that sternward velocity of water entering
the screw disc is obtained ultimately by sucking water from the
surface, it follows that if a screw is so arranged that it cannot draw
water from the surface, the sternward velocity of the water entering
the screw disc is reduced. The suction in such cases, not being
absorbed by giving velocity to the water, is likely to be exerted
upon the ship and cause abnormal thrust deduction. Once the
water has reached the screw disc it is difficult to conceive, as pointed
out in discussing Rankine's theory, how it can be given much addi-
tional sternward velocity. We must conclude that while in the disc
the change of velocity is nearly all rotary, as in Greenhill's theory.
It is true that this involves changes in pressure, and Greenhill, on
account of the increase of pressure involved in his theory, considered
itnecessary to confine the screw disc and race by a cylinder. Green-
hillhas pointed out, however, that it is conceivable to have a defect
of pressure behind the screw at the center, the pressure increasing
as the circumference is approached until at the outside of the screw
race it is normal. It should be pointed out that, since there is

quite a defect of pressure in all the water passing into the screw
disc, its pressure while in the disc can be materially increased by
the action conceived by Greenhill without exceeding the normal
pressure of the surrounding water.

To sum up, it appears that a reasonable theory of what happens to

a particle of water which is acted on by a propeller is about as follows:
When some distance forward of the screw, it is sucked aft and in
toward the shaft axis, its pressure being reduced at the same time.
Hence, it enters the screw disc with a certain sternward velocity
and reduction of pressure. As it passes through the disc its stern-
ward velocity changed but little. It has impressed upon it a

rotary velocity and an increase of pressure, so that its pressure on

passing out of the screw disc is probably very close to normal pres-
sure again for particles near the circumference of the screw race and
still below normal for particles in the interior of the race.

9. Effect of Immersion upon Suction and Efficiency. The stern-

ward velocity into the screw disc is affected by the situation of the
screw. Probably immersion alone does not affect it much. The
more deeply immersed screw is, it is true, farther from the surface
from which its water supply must come, but it is in a position to
draw upon a larger surface area. Still from this point of view there
is nothing favorable to efficiency in deep immersion, the reasons
rendering it desirable in most cases and necessary in some having
to do not with efficiency but with prevention of racing in a seaway.
If vessels worked always in smooth water, there is little doubt

that screws could be located with their tips quite close to the sur-
face, provided they did not suck air in operation, without loss of

efficiency. In fact, in a paper by W. J. Harding, read March 13,

1905, before the Institute of Marine Engineers, on The Develop-
ment of the Torpedo Boat Destroyer," we find the statement when
discussing the question of propellers of destroyers:
The least immersion of the propellers gave the best results, both
in speed and coal bill." This conclusion was deduced from con-
sideration of a number of trial results of destroyers in smooth water.
screw propeller placed under a wide flat stern, or with the flow
of water to it obstructed in any way by the hull to which it is

attached, must evidently work more after the Greenhill theory than
a screw with a free flow of water to it.

Apart from the increased thrust deduction this must involve a

reduction of propeller efficiency. It is, of course, necessary at times
to fit screws in tunnels, or so that they are hampered by the hull, but

when this must be done allowance should be made for the loss of

efficiency involved.

29. Obliquity of Shafts and of Water Flow

Shaft Deviations, Actual and Virtual.

x. Propeller designs
and calculations are usually based explicitly or implicitly upon the
assumption that the propeller advances in the line of the shaft axis.
As a matter of fact, it is unusual to find a shaft which is exactly

horizontal when the propeller is working. Shafts of center screws

are in a fore and aft line, but side screw shafts generally depart in
plan from the fore and aft line.

The divergence of propeller shafts from a horizontal fore and aft

line is seldom so great that the resolved horizontal fore and aft
thrust differs materially from the axial thrust. But there is a very
serious departure from ideal conditions as regards slip of blade dur-

ing revolution. The slip angle is a small angle, as a rule, and if the
shaft axisis changed from the line of advance of the screw, the slip

angle at one part of the revolution is increased by the amount of

angular change and at another part is decreased by an equal amount.
The slip angle is a function of the slip ratio and the pitch ratio or
diameter ratio. Fig. 170 shows slip angles for the range of pitch
ratio and slip ratio found in practice.
small size of these slip angles renders it evident that shaft
deviations occurring in practice must cause the slip of a blade to

vary materially during a revolution.

2. Wake and Obliquity of Water. The variation of wake is

another perturbing factor. The slip of the blade will be greatest

where the wake is strongest. Evidently a virtual deviation of shaft
axis can be imagined which would give practically the same effect
as the variation of wake. Finally, the water itself has a motion
across the shaft axis.
Variation of Slip.
3. The net result is that, in practice, instead
of the thrust, torque and efficiency of a blade remaining constant

during a revolution, they vary throughout the whole revolution.

To fix our ideas, suppose we consider a starboard side propeller turn-
ing outward. In considering shaft inclination we will always take
it as we proceed forward from the propeller.

If the shaft inclines upward from the propeller, the slip angle will
be decreased by the amount of shaft inclination for a blade in a
horizontal position inboard and increased by the same amount for
the blade in a horizontal position outboard. For the blade at the

top and bottom of its path there will be no appreciable change.

Similarly, for a shaft inclined inboard, as we go forward there will
be no effect for the horizontal position of the blades, a maximum
increase of slip for the top position of the blade and a maximum
decrease for the lower position of the blade.
If the wake is strongest next the hull on a horizontal line, the
result is equivalent to a downward inclination of the shaft, hence
we may say that such a wake causes a virtual downward inclina-
tion. Similarly, a wake strongest nearest the surface gives a virtual
inclination inward. Water rising up gives a virtual upward inclina-
tion, and water closing in gives a virtual inward inclination.
The table below gives the positions for maximum and minimum

slip of blades due to shaft inclination. Of course, when the shaft

has both horizontal and vertical inclination, the positions of maxi-
mum and minimum slip are neither horizontal nor vertical. In all
cases, the plane of zero effect is that including the shaft axis and
the line of advance of the center of the propeller. The plane of
maximum effect is that through the shaft axis perpendicular to the



" "
In the above, in means that the blade is in the horizontal
position next the ship. Out " means that the blade is in the hori-
zontal position away from the ship. P means horizontal position
to port for center screw and 6" the horizontal position to starboard.
" " " "
Up means blade vertical upward, down means blade vertical
The following table gives virtual inclinations of shaft correspond-
ing to wake and transverse motions of the water:



Motion of Water.

dent from what has been said that with inturning screws the shafts
should incline downward from the screws and for outturning screws
the shafts should incline outwards.
Outturning screws, with shafts inclining outward, are desirable
for maneuvering purposes.
5. Effect upon Efficiency. This question of desirable shaft
angles is of importance in practice, and it is to be hoped that some
day it will be given accurate experimental investigation. At present
we can deal with it in quantitative fashion only. As regards effi-

ciency, a moderate variation of slip during the revolution of a blade

will not seriously reduce efficiency so long as the average slip is that

corresponding to good efficiency and the variation of slip is not

extreme. But it is difficult to see how a shaft inclination as great
as ten degrees, which has been often fitted on motor boats, can fail
to be accompanied by a loss of efficiency. With such an angle of
inclination it is evident, from Fig. 170, that each blade will work
with negative nominal slip at one portion of its revolution and with
excessive nominal slip at another portion even if the average

slip is that corresponding to good efficiency.

If the thrust of a propeller were due solely to the action of the

face such a variation of slip would be wholly inadmissible. Irregu-

lar turning forces and thrust would cause serious vibration and there
would be great loss of efficiency. But the back of the blade through
its suction is always an important and often a dominant factor in
the production of thrust. The slip angle for the following portion
of the blade is greater than the slip angle for the face by the value
of the edge angle at the following edge. This edge angle is seldom
less than twelve or fifteen degrees and is often twenty-five or thirty.

Hence a shaft inclination of two or three degrees will affect com-

paratively slightly the action of the blade back, and even the large
inclination of ten degrees will seldom cause the suction of the back
to be reversed into negative thrust at any portion of the revolution.
Such a large deflection, however, is liable to produce very irregular
6. Vibration. An important consideration in this connection
is that of vibration. With turbine propelled vessels, practically
all vibration which is quite strong in some turbine steamers

is due to pounding of the water against the hull as the blades pass
or to unbalanced propeller action. There can be no doubt that the
latter cause of vibration, which is practically the only cause if the

propeller tips are not too close to the hull, is affected by the shaft

angles, and particularly advisable with turbine steamers to

it is

choose shaft which will tend to uniformity of propeller action.


Suppose, for instance, we have a propeller shaft carried by a nearly

horizontal web. We
have seen that there will be a very strong wake
above the web and vertical motion of the water will be interfered
with. In such a case, for inturning screws, the shaft should incline
down and out, and for outturning screws, up and out.

7. Obliquity of Flow. While the wake through its variation

of strength over the propeller disc produces a virtual shaft devia-

tion, it is evident from consideration of Figs. 50 to 59, showing lines

of flow over models, that the water closing in and rising up aft fol-
lows lines which will in many cases make material angles with the
shafts. The effect of the obliquity of the water flow will vary a good
deal with the position of the propeller.
For vessels of usual type it would seem that the farther aft the
propeller the less the obliquity of the water flow. But experiments
with the model of a four-screw battleship indicated that at the for-
ward screws the water was rising at an average angle of about 10
and closing in at an average angle of about 5. For the after screws
these angles were 1 1 and 4 respectively. The after screws, however,
were not very far aft. These angles seem large when we compare
them with the slip angles to be expected in practice. The obliq-
uity of horizontal water flow will usually be greater over the upper
portion of the propeller disc than over the lower, so that the virtual
wake to which the obliquity of motion is equivalent will be stronger
over the upper portions of the screw disc. Now this virtual wake
will, for outturning screws, be positive over the upper part of the
disc and negative over the lower. Fcr inturning screws the virtual
wake be negative over the upper portion of the screw disc and

positive over the lower.

The strength of the virtual wake being in the upper part of the
disc, where it is positive for outturning screws and negative for in-
turning, it would seem that side screws well forward of the stern

post should be outturning in order to make the most of the virtual

wake due to the obliquity of the water motion.

30. Strength of Propeller Blades

In view of the importance of blade thickness in many cases it is
advisable to make a careful inquiry into the matter and endeavor
to reduce to rule the stressesupon propeller blades. This can be
accomplished only by certain assumptions, which will be pointed
out and justified as they are made. In order to apply the well-
known formula for beam stress to a propeller blade, it will be assumed
that the section of a blade by a cylinder at a given radius is devel-
oped into a plane tangent to the cylinder. This section will then be
treated as a beam section. This assumption probably errs on the
safe side, since the actual strength as a beam of the curved blade
would be greater than that of a developed cylindrical section of the
i . Fore and Aft Forces and Moments. In considering the
forces upon a blade it is convenient first to consider separately fore
and aft forces, or thrust and transverse forces, producing torque. It
isconvenient to use the disc theory or Rankine's theory, by which
the thrust upon a blade may be taken to vary radially directly as
the distance from the shaft center. For a ring of water one inch
thick at ten feet radius, say, would contain twice as much water
as a ring of the same thickness at five feet radius. If each ring be

given the same sternward velocity, involving the same thrust per
pound of water acted upon, then the thrust from the ring of ten
feet radius would be double that from the ring of five feet radius.
Put into symbols, if dT denote elementary thrust from a ring of
thickness dr at radius r, we have dT= krdr where k is a constant
coefficient over the blade depending upon the total thrust. Then
integrating we have for thrust,

T= \ kr\

- where di diameter of hub and d diam-

Applying the limits >
is is
2 2

eter of propeller, we have, if T is total thrust of one blade,

1 o=
k/d* dA

This enables us to determine k, since from the above

fa = ""

Suppose, now, we wish to determine the thrust T\ from the tip to

a radius r\,

We have Ti**-\--rA

A_ '

4 4

From the above, if fi = ,

then J"i = T Q and,
if ri = -
2 2

TI = o as it should. Now we need to know not only the thrust

on the blade beyond any radius, but its moment at the radius.
The moment at radius r\ of the elementary thrust dT at radius r
is dT (r ri) = kr (r r\) dr.
Call dMi the moment of this elementary thrust. Then
, s 8 To ,

2 r,

Upon reduction we have

At the root section r\ Substituting and reducing, we have

at the root section,

A /J
6 (tf +,

hrust were concentrated at a point ki - out.

Suppose, now, the thrust
Then we should have

Whence, equating these two values of MI, we have

2 2
, d di 2 d ddi dj
i ^ 73 I j \

Upon reduction this gives us

d(d + <f

The value of k\ in theabove formula depends only upon the ratio

between di, the diameter of hub, and d, the diameter of propeller.
Numerical values are given below:
d 2 d d 4J
= 3
10 10 10 10

.689 .713 .743

These values of ki agree very well with values deduced by entirely

different methods upon the blade theory or Froude's theory. Upon
the blade theory ki is nearly constant at .7.

2. Transverse Forces and Moments. Let us now take up the

transverse moment, which denote by 2 Let dQ denote the ele- M .

mentary transverse force in pounds at radius r. Let p denote pitch

in feet, s the slip ratio and e the efficiency of the elementary portion
of the blade at radius r. Then the gross work done by the element
of the blade in one revolution is in foot-pounds dQ X 2 irr.

The useful work is

X X = dT X p - = -

dQ 2wr

-^ =

kp -

s) krdrp (i s).

Now over a blade the quantities on the right in the above equation
are constant except e. The variation in e over the part of the

blade that does the most work is probably not great, so let us assume

it constant and write = where g is a constant coefficient to be


We have seen that dQ X 2 -n-r = element of work done in one revo-
lution in foot-pounds. Then / dQ 2 ?rr = work per blade per revo-

lution = 33 -> where PI is the power absorbed by one blade.


- = / dQ 2 irr = I g 2 TIT dr = 2

= TT-e (d?-
, , ,>.
z = 4 X ^
33000 Pi I
di ) or s. ~z rr
= 42,017
a __^_

Then for Af 2 the transverse moment at any radius, r\, due

to the
moments of the elementary transverse forces from the tip in to the

radius, r\, we have


/ry2 (r rO dr = g\

Then, upon substituting its value for g and reducing, we obtain

Now as to the radial position of the transverse center of effort we

have the total transverse force equal to

The arm of this force beyond r\ is obtained by dividing moment

by total force and equals

d 2\2 y
- - /

The center of transverse effort, then, is by this method halfway be-

tween the tip and the radius considered. So if k z - denote the dis-

tance of the center of effort of the whole blade from the center of
the propeller d\, denoting the root diameter, we have

2 2 2 \2 2/

OF kz= ~

This gives us the values below of k 2 for the values of indicated,

-=.i .2 .3 4
2 =-55 -6 .65 .7

These compare fairly well with values of k z deduced by entirely

different and more complicated methods upon the blade theory.
These values of k% varied from .710 for a coarse pitch ratio of 2
to .600 for a pitch ratio of i.

Let us now recapitulate the results to this point.

Let d denote the diameter of the propeller in feet,
di the diameter of the hub or diameter to root section,

ri the radius to the point at which we wish to determine thick-

TO whole thrust of the single blade in pounds,
PI horse-power absorbed by the single blade,
R revolutions per minute,
M i fore and aft bending moment at radius r\ in Ib.-ft.

Mz transverse bending moment at radius r\ in Ib.-ft.

Then we have deduced M 1= ^ (d +r
% (d
2 Tl}

3.Moments Parallel and Perpendicular to the Sections. These

moments above are of fixed direction independent of the angle of
the section. This angle varies with the radius of the section. The
next step is then obviously, to resolve the above moments parallel
and perpendicular to the section. For the ordinary screw whose

driving face is a true helicoid this face develops into a straight line,
and we will resolve the moments parallel and perpendicular to this
line. For sections of varying pitch we will resolve parallel to the

tangent at the center of the face. Figure 238 shows an ordinary

ogival type of section expanded from its cylindrical shape. Let
6 denote the pitch angle, or the angle which the face line makes
with a plane perpendicular to the shaft. Then OB = Mi = fore
and aftmoment and OA = Af 2 = transverse moment. If c denote
. M
the resultant moment perpendicular to the face, we have from

Fig. 238,
M = OC + OD
= Mi cos + M z sin 0.

Similarly, if MI denote the moment parallel to the face, we have

MI= BD - AC = Mi sin - M z cos 0.

Now the pitch angle, depends upon the pitch and the radius.

If p denote the pitch and r\ the radius, we have

2 irr\ d 2

Denote the pitch ratio proper, or
^ by a.

Then tan =

ad 2 irr\

sm = ;
cos = =
4 Tr a 4 TT
We have seen above that M c
= MI cos + M% sin 0.

Substituting their values obtained above for MI, M 2, cos and

sin and reducing the results, we obtain
(d 2 ri) f2 TT

Let us next express r\ and d\ as fractions of the diameter d, the main

dimension. Write r\= and d\= cd. Upon reducing we have


,, =

Va + ^

2 2

T md ,
(2 + m)
v .

+ 5252 aPi~\

Proceeding in practically the same manner we obtain

- c
)Va2 + A 2
(2 + m) 525 2^R

In any particular case of design we will know P\ and R, but

generally will not know T . The equation connecting TO and PI is

= er\.

Since p = ad, this gives

= 33000 eP l
adR(i -s)
Now in practical cases e approximates but is generally somewhat
less than i j for practical slips, being greater than is only for

very high slips. So if we assume e = i s, the result will be to

make the value of T
generally greater than the truth. In other
words, we shall generally introduce a moderate error on the safe side

and simplify our expressions enormously. So write T = "4 ^ .

Also introduce the factor 12 in the expressions for c and MI, so M
that these moments, heretofore expressed in pound-feet, will be
expressed in inch-pound units. This is desirable because it is con-
venient to measure dimensions of the propeller sections in inches.

Then substituting in the expression for M c and MI, T = ^^ooo P\

and multiplying by 12 we have after reduction

= 63 ' 24

MI= 132,0007 -

- m}

+ ir*m*\-
i m~\Pi
p x
In the above expressions is the factor depending upon the work
done. The complicated fractions involve m the fraction of the
radius; a, the extreme pitch ratio; and c, the ratio between diameter

of propeller and diameter of hub. Hence these complicated frac-

tions can be calculated and plotted once for all. So write
- m)* I" m(2+m)
c= (i "I

63,024 7-
2 /-rr-T-i 3 29'

+ aJ,


(I c*-) \/a? _(- ^2^2 L

L= 132,000 -
= f
- ).
(i c ) Va2 + 7r
w 2

Then finally, M =C^,

K M^L^-
Figure 239 shows curves of C and L plotted upon for various values m
of a. For these curves c was taken uniformly at f Even if the .

hub has a generally an amply close approx-

different diameter, this is

imation for practical purposes. Since, however, for very large hubs
a correction may be needed, there is given in Fig. 240 a Curve of
Correction Factors for hub diameters. It is seen that for m= f
the factor is unity. For smaller hubs the factor is less than unity,
and for larger hubs greater than unity. Unless, however, the hub
diameter is one-third of the propeller diameter or more, it is not
worth while to undertake to correct the regular values of C and L in
Fig. 239, namely, those for the hub diameter y the propeller diameter.
Resisting Moments of Section.
4. The above expressions en-
able us, by the use of Fig. 239, to obtain very readily with sufficient

approximation the longitudinal and transverse bending moments at

any section of a given propeller of known power and revolutions. It
is next in order to consider the resistance of the section, using, as
already stated, the developed section. Referring to Fig. 241, let A B,
the length of a section in inches, be denoted by /, and CD, its thick-
ness at the center in inches, be denoted by /. The center of gravity
will be found on CD at a point G, say. Denote DG by gt, where g
is a coefficient. Let 7C or the moment of inertia about a horizontal

axis through G, be denoted by k c lP and // or the moment of inertia

about CD, by kj?t. Then for the type of section above we have
due to c M :

Tension at A and B = f-M c

fc -j-
K c lt K It

Compression at C= L ^ M c

Due to MI we have, if B is leading edge,

Tension at A= compression at B= = -
k/t 2ki Pi

These are general expressions. The coefficients g, k c and k depend


upon the type of section, and / and / are the dimensions. It will

be well, then, to consider the range of values of the coefficients g,

k c and k for various possible types of section. The most usual type

of section is the ogival, where AB a straight line and the curve


ABC the arc of a circle. This type of section, however, is difficult

to reduce to rule, the coefficients varying with the proportions. The

ogival section, however, is practically the same as a section with a

parabolic back, so the latter may be considered.
In addition to the parabolic back as representing the ordinary
type of blade section we will consider two other types of blade of
the same maximum
thickness. In one the parabola is replaced by
a curve of In the other, thickness is equally distributed be-

tween face and back, each being a curve of sines. Figure 242 shows
the three types of blade section, and below each are given the equa-
tion characterizing it, the expressions giving the area in terms of

length and thickness, and the value of the coefficients for it.
Then for the three types of blade section we have with sufficient


and the only one that need be considered in the case of material
that as strong in tension as in compression.

It would seem, then, to be the best plan in practice to design the

blade thickness from considerations of compression and then deter-
mine tension of the blade thus designed. In the rare cases where
the tension found too high it is easy to make the necessary

changes. The formula for compression at the center of the back of

the blade in pounds per square inch is,

Compression = -~
kc It

Now M = Cp

and it is seen from Fig. 242 that for all these types

of blade a safe value for ^ is 14. Then our final formula is,
Maximum Compression at Center of Back in Pounds per Square
= P i
Inch, 14 C X ^ )
where C is obtained from Fig. 239.
/V 1 1

We are now in a position to investigate the stress, not only at a

root section, but at any point along the radius, by the aid of the
above formula and Fig. 239. The result for a blade of rather wide
tips and a mean width .2 is shown in Fig. 243.
ratio of This shows
for various pitch ratios, and plotted on fractions of radius, curves of
thickness in center for constant compressive stress, the thickness

being expressed always as a fraction of the thickness at .2 the radius.

Beyond .2 of the radius these curves are so close together for the
various pitch ratios that it is impossible to plot them separately.
Below .2 of the radius the curves separate. It is seen that the outer
portion of the thickness curve in Fig. 243 is not quite straight, being
slightly curved. The curvature is so slight, however, that if we
follow the nearly universal practice of making the back of the blade

straight radially, the thickness at the tip being not zero but the
minimum that can be conveniently cast, the stress per square inch
be practically constant. Unfortunately it is clearly unsafe to

make the line of the blade back concave as we go out, thus decreas-
ing thickness and gaining efficiency for high speed propellers. It is
true that sometimes the line of blade back is made concave when the

blade has a small hub and is narrow close to the hub, but this is due
to a thickening of the inner part of the blade not a thinning of
the outer part. The only practicable method, then, of accom-
plishing reduction of blade thickness is to use material capable of
standing high stress.

Figure 243 indicates that for propellers with small hubs less
than .2 the diameter the thickness should be determined, not at
the root, but at the radius, the straight line of back being

extended inward to the hub.

For conveniencein design work Fig. 244 has been prepared. This
gives values of C, from .2 the radius to .4 the radius for pitch ratios,
from .8 to 2.0, thus covering the practical field.
6. Tensile Stresses. Coming back now to the question of ten-

sion, it seems that sections of Type 3 are the simplest. The maxi-
mum tension for it is the same as the maximum compression. But
sections of Type3 are not desirable for use. For sections of Types
i and 2 the case is not so simple. Taking the maximum tension as
that at A and the maximum compression as that at C, and denoting
by /i the tension factor or value of maximum tension -i- maximum
compression, we have
g M c i MJ
k if 2 k n

Now ~ = and with sufficient approximation we have from

M c C

Fig. 242 g =.4 and ^ = .7 1 kc .

after simplifying, = .666 + - -

L and C are
Whence, /i 1.17 given
c/ i>

in Fig. 239, but to facilitate computation Fig. 245 gives curves of

1.17 from m= .1 to m = .4, and for final pitch ratios from .6 to 2.


This covers the practical ground. For narrow cast-iron blades with
solidand hence small hubs it will generally be necessary to determine
tensile stress with care.
7. Stresses Due Centrifugal Force.
to In addition to the
stresses upon a propeller blade due to thrust and torque, there are

stresses due to centrifugal force. These are appreciable. In any

given case they can be calculated with sufficient approximation with-
out serious difficulty. If W
denote the weight of that portion of a
propeller blade outside of the radius r\ of a given section, r 2 the
radius of the center of gravity of the portion of blade and v the
circumferential velocity of the center of gravity, while g, as usual,
denotes the acceleration due to gravity, then the centrifugal force
of the portion of blade may be taken as equivalent to a single
force perpendicular to the shaft through the center of gravity of
the portion of blade. The amount of the force in pounds will be

equal to
W V2
the force and
Knowing its line of application, the
g rz

stresses upon the bounding section of the portion of blade can be

determined by well-known methods of applied mechanics.
It appears advisable, however, to make a general mathematical

investigation of a case sufficiently simple to admit of such investi-

gation and sufficiently resembling the cases of actual propellers to
enable us to apply the results of the mathematical investigation, in
a qualitative way at least, to actual propellers. It will be seen that
we can thus learn a good deal about the laws governing the stresses
of propeller blades caused by centrifugal action.
Fig. 246 shows an elliptical expanded blade touching the axis at O.
Consider the weight of each section such as CD concentrated at
the blade center line at E. Let bd denote the minor axis BN, d
being the propeller diameter and b a fraction. The equation of the
ellipse referred to the point where it touches the axis, is

y b V2 dr 4 r

where r denotes radius and y the semibreadth at radius r,

Then Breadth = 2 y = 2 b v/2 dr 4 r


Now for r substitute m - where m is fraction of whole radius varying


from o at to i at A .

Then Breadth = 2 bd Vm - m z

When we come to thickness, the axial thickness is rd, where T is

blade thickness fraction. The tip thickness is not fixed by consid-


erations of strength, being from considerations of castings, etc.


usually materially thicker than it need be for strength. wish in We

considering centrifugal force to be sure we take the tip thick enough,
so will assume it as .15 the axial thickness. It will usually be less
in practice for large propellers. Then the back of the blade center
being a straight line, the thickness at is rd ( i .85 m). m
Assuming the section as parabolic, the area of a section =
width X thickness = X 2 bd Vm m?X rd (i -.85 m}

(i .85 m) vm '
m 2

We are now able to formulate the elements of curves to be plotted

upon m
and integrated graphically to obtain the results needed.
The element of blade volume = Area of section X dr.

r = md ?
dr = -d dm.
2 2
Element of volume = -
(i .85 w) \/m m 2

Let 8 denote weight per cubic foot of the material of the blade

Element of weight = - - m) \/m m 2

(i .85)
O n o T

Element of centrifugal force = X weight =- - X weight

g *g



m (i


m) \/m
.85 ?w)
v/ w

m dm
mr, dm.


Then total centrifugal force from the tip to the section m is

8rbd4 , / x
If there is no rake the effect of the centrifugal force is simply to

cause a tension over the area. This tension =


3 g
4 rbd (i .85 m) "vm m 2

in pounds per square foot.

4g (i .85 m) Vm m

Expressed in pounds per square inch it is T ] of the stress in pounds

per square foot. So we have Tension in pounds per square inch
due to centrifugal force when there is no rake
(m) u2 8d? ,

-.S$m)Vm - m

576 g(i 57

It appears, then, that for a blade without rake the tensile stress due
to centrifugal force varies as the weight per cubic foot of the blade
material, as (cod) or as the square of the tip velocity, and as </> 2 (m)

where <
2 (#0 is a quantity depending upon radial position, blade
shape, proportions, etc., but independent of size and pitch.
Since it is usually more convenient to express angular velocity by
the revolutions per minute, denoted by R, we may substitute
for co. Also, in order to avoid small decimal factors, multiply numer-
ator and denominator by 1,000,000. Then Tension in pounds per
square inch due to centrifugal force when there is no rake
22 2
X - 8d - .

X - - .
f v

02 O),
3600 57^ g loooooo
8d2 R 2
[4000000 TT
ioooooo_ 3600 X 576 g
8d?R 2 f 4000000?^ 0i (m)
1000000 [3600 X 576 g (x
_ .85 m) Vm - m

I 000000

Figure 247 shows curves of <i(w) and (j> t

. It is seen that <i(w),
which is proportional to total centrifugal force, increases always from
tip to axis, as might be expected. Since the assumed blade has no
area at the axis, which is proportional to the stress per square inch,
(f> t ,

is infinity at the axis but falls off very rapidly at first as we go out.
We wish mainly, however, to investigate the effect of rake or
inclination upon the stresses on propeller blades due to centrifugal
action. Let id denote the total rake of the blade along its center
line, where i is a comparatively small fraction, and assume the weight
of the section concentrated at the center line. Then idm denotes
the rake from the axis to the radius corresponding to m.

Suppose we wish to determine the moment due to centrifugal

force about the section corresponding to mi.
The element of force at beyond mi is, as before, m
m (i .85 m) vm m 2
Its lever to radius mi is id (m mi). Hence element of moment


w) v w
( N ( >.
/ o ,

(m mi) m(i .85 m- dm.

O o

Moment to mi =

3rta/5 r


m (i
rm '



.S^m)\/ m
v /

m dm2
m* dm

= ,

The second integral is <i(w), but we can denote the whole thing

by </>a (mi) and after obtaining results by graphic integration use m

instead of m\. Then we have

Moment from tip to m= -

3 o
</> 3 (m) = M f
, say.

Now the moment MI is in the plane through the axis and the
center line of the blade. upon the section is best ascer-
Its effect
tained by resolving it parallel and perpendicular to the section.

If 6 be the pitch angle at radius r, tan 6 = -* = *. = -

2 irr Trmd irm

we use a to denote the pitch ratio *


Then sin 6 = cos 6 =

If Me' and ML
denote the moments resolved perpendicular and
parallel to the blade face we have

nf =
Me r
*f cos a
6 =- 7T03
, /

ML' = M '
sin 6 = - <
3 (m) _.

Finally, by applying at the center of the section forces equal and

opposite to the forces producing the moments, we have the section


affected by a force and two couples. The force is the same as the
outward force when there is no rake. The couples are perpendicu-
lar and parallel to the section and their moments are given above.
The result of the force and couples is as follows, reference being
had to Fig. 241, where B is the leading edge:
1. The force causes a certain tension over the whole section.

2. The perpendicular couple causes compression at C and ten-

sion at A and B.
3. The parallel couple causes tension at A and compression at B.
Now from consideration of thrust and torque only we have already
found that the maximum compression is at C and the maximum
tension at A Centrifugal action evidently increases the tension at

A more than at B. Hence, as regards tension we need consider the

action at A only.
As regards compression, when we neglect centrifugal action this
is a maximum at C. The tension due to the force decreases com-
pression at B and C equal amounts. Then the parallel moment
increases compression at B and the perpendicular moment increases

compression at C. We need to find which increase is the greater,

if C has
and greater compression from centrifugal action we need
consider C only.
The necessary coefficients for the parabolic sections are found in
Fig. 242. Consider the tension increases at A first. We have three

Due to force alone in pounds per square inch, fa (m).

576 g
Due to perpendicular moment, -^ -r-^- ,
where the factor 12

has been introduced because we wish stresses per square inch and
Me was calculated in pound-foot units.
Now in feet / = 2 bd ^m m?= 24 bd \/m mz in inches.
Also t = rd (i .85 m) in feet
= 12 rd (i .85 m) in inches.
So the tension per square inch at A due to the perpendicular mo-
ment is

105 Mr'
24 X 144 b^d? Vm m 2
(i .85 w)

Substituting the value of Me'
Tension at A due to perpendicular

1152 X 3 gbrW Vm-m

35 TT i
35 Tru?dTbid

uPdcPfa (m)

<j> 3




-.85 m)
Va +2

3456 g T Vm - m 2
- .85 m)
Va +

Due to parallel moment

Tension at A= 15
-* X -12 ML'

Reducing this similarly we have

Tension at A due to parallel moment

576 g b m-W ( 2) ( z
_ g5

Suppose now we denote by N the tension per square inch due

to centrifugal force only and express these other tensions in terms
of N:
We have

576 g 57 6 g (J
- -85
o w -
) (

Then Tension at A due to perpendicular moment

6 r ^(w) -.

_ __
= -^-^>4 .

Tension at A due to parallel moment
,, i 03 (m) a
~ 5 ~ j\j-
b fa (m)
- '

In the above ^ and 0s involve a as well as m and should be expressed

by contour diagrams.
Consider now the compression at C due to the perpendicular


-^ s2
12 X Me As
before in inches

I = 24 bd (m w )*,

t = i2rd(i .85 m).


Hence compression
13.125 X 12 Me
24 X 144 X br
d3 (m -w) 2 4
(i -.85 m)'

13.125 jr
u?8rbid5 <f)3 (m) m
288 3 g br d?
(m - w )* (i - .85 m)*Va*+v*ni<

26.25 E2 i 03 (m) m
576 g 3 r
- W2)i (i _ .85 w)
vV+ 2

And in terms of N
Compression at C
() (i -.{

We can express the ratios between extra compression at C

and compression at B due to parallel moment. The latter is the
same as tension at A due to parallel moment
(m) (
Extra Compression at C
moment Compression at B
Parallel 15 T a i.&$m

Now we may safely say that in practice b is greater than 47.

If we put b = 4 T, TT = V we nave f r above ratio,>

m (m
w )*

a i .85 m

The hub is such that m may be taken as .2 or more. Putting

w =.2 we have ratio above = So for propellers in practice

the extra compression at C due to centrifugal action will always be

greater than that at B due to the parallel moment. When, too, we

recollect that there is a large opposing tension at B due to the per-
pendicular moment, it is obvious that the maximum compression is
at C, and only that need be considered.
Figures 248 and 249 show contours of 4 and 5 plotted on a and </>

m and curves of $i(w), <& and 4>z(m] which involve m only are
shown in Fig. 247.
We have finally for stresses due to centrifugal forces

Tension in pounds per square inch neglecting rake = N


Extra compression at center of blade back = AM- <

dd?R 2 ^ (i ^
= <p t [- 94 i
m pounds per square inch.
1000000 \T I
i i \

9^ + 7* 95 + I )
3 T I
dd?R /2 i i, \.
= 9,
, .

9<+ r 95+ 1 m pounds per square inch. . ,

1000000 \3 T b I

In the above i is ratio between rake and diameter, T is ratio

between axial thickness and diameter, and b is ratio between maxi-
mum blade width and diameter and may be taken as 1.188 (mean
width ratio).
The above formulae and the accompanying figures apply strictly

only to blades whose expansion is an ellipse touching the axis and

whose tip thickness is .15 the axial thickness.

The methods used can be followed to determine 9i(w), (f> ^>z(m), t

4>4and 5 < for blades of any type, but the results of Figs. 247, 248
and 249 can be applied in practice with sufficient approximation
to any oval blade that does not depart widely from the elliptical
Since centrifugal stresses increase as the square of the tip speed,
they evidently need to be given much more careful consideration
for quick running propellers than for those of moderate speed. Thus,

suppose we had a manganese bronze propeller for which dR = 4000,

or the tip speed is over 12,000 feet per minute. For manganese
bronze 5 = 525 about. Then N = 525 X 169^ = 84009^. For
m = .3, <f>i= .135 about, soN = 1134. If the pitch ratio is about

94 =
- has the value of
unity, 2\ about, and if 3 or the rake is three
times the axial thickness, -941 = 6, or increase in compressive
stress at .3 the radius is the largeamount of 6700 Ibs. This is an
extreme but not impossible case. As tip speed falls off, stresses due
to centrifugal force decrease rapidly, but it would seem the part of
wisdom to avoid them entirely by avoiding backward rake. More-

aver, it seems advisable when tip speed is very high to give a moderate
forward or negative rake, thus opposing the tensile and compressive
stresses due to the work done by opposite due to the cen-

trifugal forces. When would add to the

backing, centrifugal force
natural stresses, but propellers are not worked backward at maxi-
mum speed.
In calculating stresses due to centrifugal force we need values of
5 or weight per cubic foot of the various materials used for pro-

peller blades. For manganese bronze or composition we may use

525 for 5, for cast iron 450 and for cast steel 475.
8. Stresses Allowable in Practice. While for quick-running
propellers centrifugal stresses must be calculated separately, in the
majority of cases they are not very serious and may be allowed for
by using a low stress in our main strength formulae.

=S = P 1 i
Compressive stress in Ibs. per sq. in. c 14 C X %'
1\ LI

Tensile stress in Ibs. in. = ST = S + - -

per sq. c (.666 1.17)
\^ L

In applying these formulae to the root section of any blade we will

know C, PI, R and /. Then we fix t by giving Sc a suitable value
and calculate ST to see if that has a suitable value. Now what
are suitable values of Sc for the various materials of which we make
propeller blades ? They cannot be fixed arbitrarily from considera-
tion of only the tensile and compressive strengths of the material.
For one thing our formulae are approximations only. In order to
apply the methods of Applied Mechanics we start by developing the
cylindrical section of the blade into an ideal plane section. It is

probable that this ideal section is materially weaker than the actual
section, especially in the case of propellers of varying pitch.
if were the only perturbing factor, we could allow high stresses
in the formulae, because the stresses per formulae would be greater
than the true stresses. But when we consider the conditions of

operation of propellers we find other very serious perturbing factors

which we cannot reduce to rule. In the formula, PI is the average
power absorbed by the blade. But even in still water the blade,
owing to inequalities of wake, will absorb more power than the

average at one portion of the revolution and less at another. And

in disturbed water, what with the motion of the water and the
pitching of the ship, the blade is liable to encounter stresses very
much in excess of those due the average power which it absorbs.
This is especially likely to be true of turbine driven propellers.
With reciprocating engines, when a propeller encounters abnormal
resistance the engine will soon slow down, the kinetic energy of the

moving parts being rapidly absorbed. With turbines, however, we

are likely to have the kinetic energy of the moving parts per square
foot of disc area much greater than for reciprocating engines, and
the flywheel action, so to speak, of the moving parts is then capable
of causing a relatively greater extra stress.
To determine with scientific accuracy allowable stresses for use
in the formula we would probably have to test to destruction full-
sized propellers which is impracticable. The next best thing is
to find from the formula the stresses shown by actual propellers
which have been successful in service, and also those of propellers
which have shown weakness in service. We can thus establish, with
accuracy for practical purposes, the maximum stresses that
can be tolerated. The advantage in this connection of a formula
upon a sound theoretical basis is that a stress found satisfactory for
a fine-pitched, quick-running propeller, for instance, will be almost
equally satisfactory for a coarse-pitched propeller, and vice versa, so
that satisfactory allowable stresses can be deduced from less data
than would be necessary for a formula partaking largely of the rule
of thumb nature.
There are advantages in the use of a simple semi-graphic method
which will enable data from completed vessels to be recorded for
use in design work.
We have deduced as the final formula for S c the compressive
stress in pounds per square inch for blades of the usual ogival section

where C is a coefficient depending on radius and pitch ratio, PI is

the power absorbed by the blade, R denotes revolutions per minute
of the propeller and I and t are width and thickness respectively of

the blade in inches. Also we should generally use in determining

Sc the values of C, I and t at about .2 the radius of the propeller.
Let us now express I and t in terms of coefficients and ratios already
Put / = 1 2 chd where d is diameter in feet, h is mean width ratio
and c is a coefficient depending upon the shape of the blade.
It is not such a simple matter to determine a rigorous expression
for /, because the tip thickness is more or less independent of the
root thickness.
If rd denote axial thickness as usual, and krd the tip thickness,
we have for .2 the radius

t = 12 rd [k +.8 (i
- k)]= 12 rd (.8 +.2 k}.

In practice k is seldom much less than .1 or greater than .2

Now k = o, t = 9.6 rd, k =.i, t = g.&4.rd, k =.2, t = 10.08 rd.

So it is a sufficient approximation for practical purposes to put

t = 10 rd.

So, returning to the stress formula, we have

14 C PI i
OC*c = 14 C
^ v/
X 2 2
12 chd IOOr J I2OO

Let Ci= Figure 250 shows plotted upon pitch ratio a curve
1 2OO
of Ci for .2 the radius.

Then S =
^~ X ~^-
CiPt = #, =
Suppose, now, we put chrz y:

then we have 5C = -
Figure 251 shows contours of values of Sc plotted on x and y. In
the case of a given propeller we know or can readily calculate chr2 and
CD Hence, we can locate a spot on Fig. 251 corresponding to

the propeller which will show the root compression or value of Sc in

pounds per square inch. Figure 251 shows by crosses a number of
spots each of which corresponds to an actual propeller. They are

nearly all for vessels of war, and all for manganese bronze or other
strong alloy. It is desirable, when using the method for design work,
to reproduce Fig. 251 on a large scale. It is evident from Fig. 251
that the designers of the propellers referred to differed widely as to
the allowable stress. No. n
refers to a destroyer which would

very seldom develop maximum power, and then only in smooth

water. But even for such vessels it is not advisable to go to such
stresses. No. 14 was a vessel which much exceeded her designed
power, on trial, and also sprung her propeller blades. With man-
ganese bronze and similar alloys now available it is inadvisable to
exceed 15,000 Ibs. even for destroyers. For other fast men-of-war
which seldom develop full power, suitable stresses, based upon full
power, are 10,000 to 12,000 pounds per square inch. For merchant
vessels,always at nearly full speed, particularly passenger steamers
that are driven hard in rough weather, it is not advisable to exceed
5000 to 6000 Ibs.The above all refer to blades of manganese bronze
and similar alloys. Good cast-steel propellers can be given the
same stresses as those of manganese bronze.
For cast iron it is advisable not to exceed 5000 Ibs. for compression
and 2000 Ibs. for tension.
As already stated, designers differ widely as to the proper stresses
to allow for propeller blades. It is a simple matter for any designer
with an accumulation of data for actual propellers to record it on
a large diagram similar to Fig. 251 and form his own conclusions
as to the stresses which he will allow in a particular case.
While it is desirable for a designer fully to understand all de-
tailsinvolved in determining propeller blade thickness, it may be
pointed out that when centrifugal forces are not serious, and the
blade thickness from considerations of compressive
is to be fixed
stress only, Figs. 250 and 251 are all that need be consulted.
For when number of blades, diameter and pitch have been deter-
mined we can determine P\, R and d. C\ can be taken from Fig.
250, so we will know From the blade outline we can deter-

mine h and c, the latter usually falling between .6 and .8 in practice.

Thus a practical case, after having calculated ch we need only
to determine r.

* *
So we enter Fig. 251 with the value of and from the stress

chosen determine c^r 2 and ,
c/s being known r and T are readily
9. Connections of Detachable Blades. While somewhat apart
from the question of strength of propeller blades it seems advisable

to consider briefly the question of the strength of the connections

of detachable blades. We
have seen that the formulae for trans-
verse and fore and aft moments in pound-feet are:

Fore and aft moment n.
M\= To
(d+R )(d-2r }* l

- l

Transverse moment M 2 5252

^= "Tjj

Also with a margin for safety we may write TQ =

Making this substitution and multiplying by 12 to reduce moments
to inch-pounds, we have:

= P (d
-- l

- 2

Now with sufficient approximation we may write d\ d.

Also we may take r\ or the radius to hub flange to which the blade
is bolted, as \ the propeller radius with a slight error on the safe
side. Substituting and reducing, we have in round numbers
M!=- d PI K
, M
= 52,400

These two moments may be compounded into a single moment

whose direction makes with the direction of the shaft axis
~ l ^-^ -
tan and whose amount in inch-pound units is
11 60

p i . It NO
2 , /n6oV =H P!
(524) say.
Y/ +(^-yj
The amount and angle of the moment depend upon - and a only.

Figure 252 shows plotted upon a, or extreme pitch ratio, curves

of values of the angle of inclination of the moment and of the
The moment above must be resisted by the bolts securing the
blade flange to the hub and the flange itself. The bolts are, of

course, disposed on each side of the direction of the moment, and

it is good practice to use more bolts for the side where the bolts
are in tension when going ahead. Thus, if there are nine bolts in
all, five will be in tension when going ahead and four in tension
when backing.
Theoretically, the blade flange will pivot under stress about some
point on its extreme circumference and the leverage of each bolt
will be the length of a perpendicular from its center to a line drawn
through the pivoting point tangent to the circumference.
For a conventional assumption, however, which is an adequate
approximation, v^e may take the effective leverage of each bolt in
tension as the diameter of the circle through the center of the

Investigation of actual propellers upon this basis indicates 3000

pounds per square inch as a fair average of the stresses allowed on
steel flange bolts by designers, the actual stresses varying from less

than 2000 pounds to some 4000 pounds.

Even after making all allowances for the conditions of service it
would seem that 3000 pounds per square inch is a low stress for
such bolts and that 4000 pounds or more might be used without
For quick running propellers the stress taken account of should
include that due to centrifugal force upon the blade. The expres-
sion for force in pounds is

and for moment in pound-feet,

The moment may be taken as parallel to the shaft axis. It is seen

from Fig. 247 that we may, with fair approximation, use .09 for

<i(w) and .04 for ^>a(w). Substituting these values, and putting
2 irR
g = 32.16 and co = ,
we have:

Force in pounds =

Moment in pound-feet =

31. Design of Propellers

i. Number and Nearly all the matters of detail

involved in propeller design have been already considered, but it is
proposed briefly to review the general considerations involved, and
illustrate the methods already explained by working out a few ex-

amples. The question of the number and location of propellers is

not very often an open one at any stage of the design, being usually
fixed by practical or other considerations which have little to do

directly with propeller efficiency. From the point of view of pro-

peller efficiency only, the best location for a propeller is in the center

line,as far aft as possible. In the center line it gets the maximum
benefit from the wake and the farther aft it is the less the thrust
deduction. Practical considerations of protection from damage re-

quire the screw to be forward of the rudder, but a suitable arrange-

ment by which the screw was located abaft the rudder, so that its
suction would not produce appreciable thrust deduction, would un-

doubtedly increase efficiency of propulsion. Since, however, suc-

tion will have no retarding effect upon a fore and aft plane, about
the most that can be done in practice to reduce thrust deduction
upon a single screw vessel is to make the after portion as fine as
possible. In many cases there might be more done in this direction
than is done. Fineness at the water surface is what is needed.

As to vertical location, it is the usual practice to locate screws

as low as possible. For seagoing ships this is desirable to reduce

racing, and even for ships intended for smooth water service only,
it isgenerally necessary, because such vessels are usually of shallow
draft, and to get the propeller sufficiently beneath the water sur-
face it must be placed low. But propellers are not placed so low
that their tips project below the keel if this can be avoided.

This is simply to reduce risk of damage in case of grounding, and

in some cases it is necessary to ignore this risk and allow the

propeller tips to go below the keel.

There is little doubt, that contrary to what is generally supposed,
a propeller for smooth water work is more efficient the closer it is
to the surface, provided it is not so close that it draws air from the
surface. This, for the reason that in this position it gets the greatest
useful reaction from the wake. Frictional, wave, and stream line
wakes are strongest near the surface.

One is apt to conceive of the frictional wake as a vertical belt of

nearly uniform horizontal thickness. But an examination of Figs.
50 to 59, and careful observations of actual ships, would seem to
indicate that the frictional wake abreast the stern widens rapidly
as we approach the surface, and in fact we may almost regard the
wake as made up of a vertical layer close to the ship and a horizon-
tal layer extending out some distance from the ship, but not extend-

ing deeply into the water. The higher a center line propeller is the
more it gains from the vertical layer, and if it is high enough to reach
the horizontal layer it gains still more. But as already pointed out,
it isnecessary to give a good submergence to the screw of a sea-
going vessel to avoid racing in a seaway. A broken shaft is too
serious a matter to be risked in order to secure slightly greater pro-

pulsive efficiency in smooth water. Furthermore, in rough water a

deeply submerged screw which does not race will have much higher

propulsive efficiency than one close to the surface that is racing con-

stantly. So in practice we usually find screws of seagoing vessels

immersed as deeply as practicable.
best location for a side propeller is probably the nearest loca-
tion practicable to the best location for a center line propeller.
Where twin screws are fitted they would, under this rule, be placed
as far aft as possible and as close to the center line as possible.
It must be said, however, that the fore and aft location of a side
screw appears to have surprisingly little effect upon its efficiency.
We saw in considering actual and virtual shaft deviations that for
a four-screw vessel the after pair were about as badly off in this
respect as the forward pair. We would expect, however, a priori,
that a side screw well forward would usually have greater virtual

shaft deviation than one well aft,and would also gain less from the
wake and have a greater thrust deduction.
It is undesirable to place screws so that their tips are too close
to the surface of the hull. When a screw tip strikes the belt of
eddying water adjacent to the hull, the virtual blows resulting are
communicated to the ship, shaking rivets loose and causing vibra-
tion. The irregular forces upon the propellers also cause vibration
of the ship.
In some twin-screw ships this trouble has been partially avoided
by leaving an opening in the dead wood abreast the propellers.
This saves the ship, and with large propellers of moderate speed of
revolution the tips can be brought quite close to one another with-
out giving trouble. For small, quick-turning propellers, such as those
fitted with turbines, vibrations are very likely to be set up unless the
blade tips are kept well clear of the hull, say 30 inches to 36 inches.
It seems a pity to lose any of the beneficial action of the wake, and it
ispossible that if the hull abreast the propeller tip were made of cir-
cular shape, with the shaft as a center, specially strengthened to stand
the pounding, and the propeller tips fitted close to the hull so that
they caught the dead water through a large arc, the beneficial effect
of the wake might be had without very objectionable vibration,
though such propellers would probably be noisy. That is a matter,
however, which could be determined only by a full-sized trial. The
only solution now known to be successful is to keep the blade tips

well clear and accept the slightly reduced efficiency.

When triple screws are fitted, it is obviously desirable that the
races from the side screws should almost or entirely clear the disc
of the center screw. This result is best attained when the side
screws are forward of and above the center screw.
For a side screw located well forward the question of virtual
deviation due to the water rising up and closing in aft is frequently

given less attention than it should receive, resulting in loss of

efficiency and vibration from the screws.
Whenfour screws are fitted the after pair are located in the
natural location of twin screws, and the forward pair are placed
forward and higher so as to avoid interference as far as possible.
These forward screws, if badly placed, are liable to serious virtual

shaft deviations, and the questions of their location, shaft angles,

etc., should receive most careful consideration. They may, from
their high location, get a better reaction from the wake, and hence
not lose in propulsive efficiency as compared with the after screws.
The number of screws depends upon various considerations. If
there is no limit to diameter and revolutions, there is no question
that the single screw should be the most efficient. There is prob-
ably not much to choose between twin and triple screws as regards

propulsive efficiency. Quadruple screws are likely to be somewhat

the least efficient as regards location. In practice, however, in a
given case, diameter and revolutions are not unrestricted, and the
number of screws is apt to be fixed from other considerations than
those of slight differences of efficiency due to number of screws.
Twin screws were adopted for men-of-war primarily to secure
greater immunity from complete breakdown, greater protection of
screws and engines on account of smaller size, and ability to do
some maneuvering independent of the rudder. The same considera-
tions influenced the adoption of twin screws for high-class passenger

vessels, but another consideration came in here. With the very

great powers used for such vessels the engines or shafts became
too large with single screws. This consideration has also largely
influenced the adoption of triple and quadruple screws.
With the advent of the turbine the question of revolutions

already of importance in fixing the number of screws for quick-

running engines became a very important one.
For steam economy and weight saving the turbine should use
high revolutions. But a propeller which absorbs great power at
high revolutions must be given so much diameter in proportion to its
pitch that its efficiency becomes too small. Hence, with turbines
we usually find three or four shafts. In the early days of turbines
multiple screws were often fitted two or three on each shaft.
This practice has now been abandoned, however, as a result of

experience, the present practice being to fit but one screw on each
While in many cases with turbines it is desirable for the best

economy to use three screws, it is rather difficult with three screws

to secure satisfactory arrangements for the rudder post and rudder.

Still it is possible to do this, and three screws are used until ques-
tions of economy or size of units drive us to the use of four screws.
2. Direction of Rotation. Obviously, when we have a center
line screw it will give the same efficiency whether it is right-handed
or left-handed. Hence the direction of rotation of single screws and
of the center screw of triple screws is immaterial. The desirable
direction of rotation of side screws depends upon considerations of
water flow and shaft obliquity already discussed in detail.

For ships as they are, in the vast majority of cases, it seems

probable that side screws would be slightly more efficient if outward
turning. For side screws very far aft, with shafts supported by
struts, so that the fittings for carrying the shafts do not interfere
with the natural water flow, it matters little as regards efficiency
whether the screws be out turning. With shaft webs approach-
in or

ing the horizontal, the side screws should be outturning for effi-
ciency. With webs approaching the vertical, they would be
more efficient if
inturning. Such shaft webs are, however, prac-
tically unknown. Side screws materially forward of the stern, how-
ever their shafts are supported, should turn outward for the best
As regards efficiency, then, in about all practical cases side screws
should be outturning. For maneuvering by means of the screws
alone, when a vessel has not steerage way, outturning screws are
distinctly preferable for practically all types of vessesl. For many
vessels this consideration alone would outweigh minor difference
of efficiency, but as outturning screws have the advantage as re-

gards efficiency in nearly all practical cases, they should be adopted

in the vast majority of cases. Cases may occur where it is a matter
of indifference, and cases are conceivable where, as with vertical
shaft webs, inturning screws are more efficient, but outturning
screws should be the rule and inturning screws should be fitted
only for good and sufficient reasons, which in practice will exist very
seldom indeed.
3. Number When the number and location of pro-
of Blades.

pellers are settled and becomes necessary to get out finally the

we will know the power which it is expected

design of the propeller,
to absorb and the revolutions it is to make. The speed of the ship

willbe known, and we can estimate the wake factor and thus deter-
mine the speed of advance. About the first thing to be settled is the
number of blades. Two-bladed propellers are hardly worth consid-
ering for jobs of any size.Figure 217 indicates that appreciable gain
in efficiency is not to be expected from them, and they are distinctly
inferior as regards uniformity of turning moment and vibration.

So, in practice, the choice will lie between three blades and
four blades. Model experiments of a comparative nature appear to
indicate that three-bladed propellers are essentially more efficient

than four-bladed.
It is seen from Fig. 216, however, which probably exaggerates,
ifanything, the inferiority of four-bladed propellers that this inferi-
ority is small, and it may well happen in practice that a four-bladed
propeller exactly adapted to the conditions will be superior to a
three-bladed propeller not so well designed.

Many designers are firm believers in the superiority of the four-

bladed screw as well as many sea-going engineers. Probably in
rough water the four-bladed screw will show a slightly more uniform
turning moment and less tendency to produce vibration. But some
of the fastest Atlantic liners that are driven at top speed in fair
weather and foul have three-bladed screws. All things considered,
there are probably few cases in practice where with equally good

design the three-bladed propeller is not somewhat to be preferred.

It should always be lighter and cheaper, and this is a matter worthy
of consideration, especially when the propeller is to be made of an
expensive composition.
In some large four-screw turbine jobs, two of the screws have been
made four-bladed and two three-bladed with satisfactory results.
With this combination the chance of objectionable vibration due
to synchronism is practically eliminated. Where special reasons such
as this exist, or where strong prejudices exist, it may be advisable
to use four-bladed propellers, but in the vast majority of cases three
blades should be used.
We have seen in Section 25 that propellers witn solid hubs are
slightly more efficient than those with detachable blades. The dif-

ference is small, however, except for quick-running propellers, which

are usually of small diameter. There are great difficulties in the

way of accurately casting and finishing large propellers with solid

hubs say propellers over 12 feet in diameter. Hence, such pro-
pellers should nearly always be made with detachable blades.
4. Material cf Blades. For the material of propeller blades
we have a choice between cast iron, cast steel, and some copper

alloy, such as composition, manganese bronze or other special alloy.

Forged steel blades. have been used, but are not found now.
For such a vessel as a tugboat, with its wheel near the surface
and regarded as desirable.
liable to strike floating objects, cast iron is

Its brittleness and weakness here become virtues, for when a blade
strikes something it breaks without endangering the shaft or engine,
and it is cheaper and shorter to renew the propeller than the shaft
or portions of the engine. Cast steel is superior to cast iron in
strength and is largely used for merchant work.
Manganese bronze and other special alloys can now be had with
strength equal or superior to that of cast steel. They can be given
a better surface, and from the point of view of efficiency of propul-
sion are decidedly the better materials. They have two drawbacks.
The first cost is higher, and through galvanic action they are liable

to cause excessive corrosion of the portion of the ship's structure

adjacent to them. This damage can, however, be neutralized in

practice by the use of zinc plates properly secured to the hull.
A very serious objection to iron and steel blades is their tendency
to corrode. The backs of the blades where there is eddying water
probably mixed with air seem peculiarly subject to extensive and
rapid corrosion.
The practical conclusion is that noncorrosive blades should by
all means be used, unless their first cost prohibits them for the job

in hand or unless for special reasons cast iron is indicated.

But in many cases cast iron or steel blades as a gift would be in
the end more expensive than noncorrosive blades, owing to the loss
of efficiency and greater coal consumption caused by their extra
friction when corroded. This extra friction is the more objection-
able the finer the pitch of the propeller.
5. Width of Blades. The blade area of a propeller of given
diameter and pitch varies directly as the mean width ratio. While
it has sometimes been thought that comparatively small changes of

blade area had large effects upon propeller action and efficiency, this
view is hardly sustained by practical experience. When cavitation
is not present, rather large changes in blade area produce quite
small effects. It should be remembered, too, that in practice change
of blade area involves change of blade section with attendant change
of virtual pitch.
The p8 diagrams indicate clearly that when cavitation is absent
the best mean width ratio is between .25 and .30. For mean width
ratio of appreciably reduced, and for wider
.35 the efficiency is
blades still it falls off quite rapidly. These conclusions are for very
smooth blades. In practice blades become more or less roughened
and foul, and when this is the case the wider blades will have the
greater loss of efficiency.
The conclusion indicated as a practical rule is that where cavita-
tion is not to be feared the best all-round mean width ratio is about
.25 or less. To avoid cavitation wider blades up to a mean width
ratio of .35 or so should be used, even with thin blades of hollow-
backed type. In extreme cases even wider blades may be required,
in spite of their excessive friction loss.

Examples of Design.
6. The principles governing propeller
design and the application of the methods that have been given
will now be illustrated by some typical cases.

First Case. Design the propeller for a turbine Atlantic liner

which develops 80,000 shaft horse-power upon four screws making
200 revolutions per minute each and has a speed of 28 knots. Here
we may take the propeller power as 20,000. The first thing neces-
sary is to estimate the wake factor. In the case of a job of such
importance this would be done nowadays from model experiments.
Let us suppose that we are considering the after screws and that
the wake factor is 10 per cent.

Then ^= .9 X 28 = 25.2.

2OO V/ 20,OOO _
So p = r^~ =
2 5
' 8.87.
(25. 2)

Also dJ = d
* - (20,000 X..25.2)*
J '
=.2608 5.

We are now prepared to enter the p8 diagrams (Figs. 211 to 214).


Since, however, we know that this is a case where cavitation is to

be carefully provided against, we would expect to use a blade of wide
type, so we will use only Fig. 214 for a mean width ratio of .35.
In Fig. 214 for p =
8.87 the best pitch ratio is 1.140 and the best
value of 5 =
57.4. Then diameter d =
.2608 57.4
= i4'-97 and X
= 14.97 X
1.140 = if. 07, the real slip being 25.2 per cent.
These for a blade thickness fraction of .03. Now the power PI
absorbed by each blade is 6667. From Fig. 250 for a pitch ratio
of 1.14, Ci= 910 and (i4-97)
3 = 3355.

Hence ( v
for Fig. 251 x = CiPi
=-= =
X 6667
*- = 9.04.
Rd* 200 X 3355

Now it seems advisable in such a job to keep the stress down to

moderate limits. So let us try for it 7500 Ibs. per square inch.
From Fig. 251 where x
= 9.04 and compressive stress is 7500,
y = = Now we know h =.35, and if c = which
ch-r .ooi2 about. ,

will be somewhere near the truth, we have r 2 = - * = .00514;

7 =.072, axial thickness = 12". 9. Now Fig. 214 being based upon
a blade thickness fraction of .03, it is necessary to correct the
results obtained by using Fig. 215. From this figure when
p = 8.87 for each .01 increase of T the diameter should be de-
creased i.i per cent and the pitch ratio increased 0.9 per cent.
So the total decrease in diameter would be i.i X 4.2 = 4.62 per
cent and increase of pitch 0.9 X 4.2 = 3.78 per cent. This would
make the diameter 14.97 X .9538 = 14'. 28, pitch 17.07 X 1.0378
i7'.72. If we allowed a stress of 10,000 Ibs. per sq. in. which might
be admissible in such a high-class job as this we would have from
Fig. 251 y
= chr* = .ooo(). Whence, for c = h = .35,

2 - X J3
T =.0621, axial thickness = n".2.

The reduction of thickness is not very much, but we could probably

stand an axial thickness of 12 inches.

Now the tip speed will be over 9000 feet per minute and even
with the best possible shape of blade section some cavitation is to
be expected. So as much increase of slip would involve rapid fall-

ing off of efficiency, it would seem advisable to make the propeller

a little large in order to provide against this and adopt as the final
dimensions: Diameter 15 feet, pitch 17 feet 6 inches, mean width
ratio .35, axial blade thickness 12 inches. The propeller effi-

ciency to be expected, barring cavitation, is about 67 per cent.

Second Case. Design the propeller for a large twin-screw tur-
bine destroyer to make 34 knots with 25,000 shaft horse-power at
800 revolutions per minute, the wake fraction being .03.

Then VA = 34 X-97 = 32.98,

_ 800


This too is a case where cavitation is to be carefully guarded

against, so we consider only Fig. 2 14.

From this figure for p = 14.3 the best pitch ratio is 1.004 and
8 = 60.3, the propeller efficiency being about 62 per cent.

Then d = 6'. 036, p = 6'.o6.

Consider now blade thickness, PI = 4167, and from Fig. 250

Ci= 1015, alsod3 = 220.

Then from Fig. 251

Ci-Pi _
~~ ''
1015 X 4167 _
Rd? 800 X 220

This is a value of x beyond the limits of Fig. 251, but to use this
method a designer should prepare an enlarged and extended copy
of Fig. 251. In this case we wish to use a high stress, say 12,000
Ibs. It will be found that using this stress in an enlarged copy of
for x = 24, c/zr
Fig. 251 we have

In this case, too, we may put c = and we have h = .35. Then

Z = -.0020 X 3^=.00857, T = .0026.
Axial thickness = 6| ".
In this case, too, there would be a decrease of diameter of about
7 per cent and an increase of pitch of nearly 6 per cent from Fig.
215. But with a tip speed of about 15,000 feet per minute there
will almost certainly be cavitation, and it is not safe to reduce the
diameter. It does seem advisable, however, to increase the pitch

slightly to provide against excessive slip. So the dimensions indi-

cated are: Diameter 6 feet inch, pitch 6 feet 5 inches, mean width
ratio .35, axial blade thickness 6f inches. The propeller efficiency
to be expected in the absence of cavitation is about 62 per cent,
but this is a case where the actual efficiency depends largely upon
the amount of cavitation. Some cavitation is almost unavoidable.
The propeller in this case would be cast with solid hub. We thus
lose the possibility of varying the pitch and hence adjusting the
propeller to the engines after trial. In cases where there is uncer-
tainty it is possible virtually to provide for this, however, by making
the propeller originally a little large. If trials show it too large,
blade tip can be cut off to suit, being careful not to throw the pro-
peller out of balance.
Third Case Design the propeller for a twin-screw gunboat to
make 17 knots with 3700 I.H.P. at 156 revolutions per minute, the
wake fraction being .08.
Then VA= 17 X .92 = 15.64. We are dealing now with I.H.P.
and must estimate the propeller power. Assume it .9 of the I.H.P.
P= _ = I66s>

_ 165 Vi66 5 , , . g (1665 X 15-64)* _. - .
P 2 5

(I5-64) (165)'

This is a case where with proper blade section we need not seriously
apprehend cavitation. Hence we should try all four pd diagrams.
The results are tabulated below:

pd diagrams, Fig. No

pS diagram, Fig No

but this will result in increased diameter, which is already by no

means small. Also reduced revolutions are almost certain to be
objectionable as regards the engine. Another practicable method
of reducing p is to use twin screws, but this has obvious objections.
The trouble is essentially the same as encountered with moderate
speed turbine vessels, namely, that the desirable engine revolu-
tions are too high for a propeller of high efficiency. There is a
further trouble, namely, that the propeller of high efficiency may

require an impossibly large diameter. Still, the best solution of the

problem is the same as for the turbine, namely, a satisfactory speed

reduction gear of high efficiency, so that both engine and propeller

can be given the revolutions best suited to their needs.
be observed that the propeller of best efficiency has to
It will
work at a very high real slip. This essential condition is masked
in practice by the fact that the wake fraction is large, so that the

apparent slip is very much below the real slip. In fact, for such
vessels very good results may be obtained when the apparent slip
is zero.

The fact that the best we can do in such cases is to work a pro-
peller of fine pitch, and hence low maximum efficiency, at a high
slip, so that its efficiency is well below its maximum, is the main
reason for the rapid reduction of efficiency with large values of p.
For small values of p propellers can usually be worked much closer
to their point of maximum efficiency.
It will be observed that while for the .25 M.W.R. the best pitch
ratio is .891, this can be made i.i with a reduction of possible effi-

ciency from .600 to .585 only. But the diameter can be reduced
thus from i9'-47 to 17'. 95 or over 18", the pitch rising from i7'-35
to 19'. 74. If a four-bladed screw is used the diameter can be re-

duced still more.

32. Paddle Propulsion
The vast majority of sea-going vessels are propelled by screws,
and vessels using paddle wheels are practically all engaged in chan-
nel, bay, lake or river service.
i. General Features. It is obvious that a paddle wheel through
its construction and method of operation approaches more nearly
than the screw propeller the ideal frictionless propelling apparatus

discussed in Section 22. If, for instance, we regard a paddle wheel

as discharging directly astern a column of water of area equal to the
area of a paddle float and with velocity equal to the difference
between the peripheral velocity of the center of the float and the
speed of advance of the ship, and make the further assumption that
the action is frictionless and that the water is discharged without

change of pressure we have an ideal propelling instrument to which

Fig. 171 applies.
This leads us to the 'conclusion that if A denote the area of a
paddle float, V the speed of advance in knots and P the shaft
p =
horse-power absorbed by the paddle wheel <f)(e)
= (f>(s) =K r

where the coefficient K' is a function of the slip. For paddle wheels
the slip is generally reckoned with reference to the peripheral speed
of the paddle centers. If V p denote the peripheral speed of the

paddle centers in knots and V the speed of advance of the vessel

in knots,
y _y

2. Fixed Blades. -- The earliest paddle wheels had the blades

on radicallines, as indicated diagrammatically in Fig. 253, and many
paddle wheels are still of this type.

Figures 254 and 255 trace out the successive positions of a

single float with reference to water for 30 per cent slip and 10

per cent slip respectively. The direction and relative amounts of

the velocities of the inner and outer edges of the floats are also
The marked W.L. indicates a water line such that the blade

has its upper edge immersed in its deepest position about one half
of its breadth. There is of course minimum obliquity of action
when the blade is vertical, in its deepest position, and it is desira-
ble that the blade should do as much work as possible when deeply
immersed. That would require it to enter the water edgewise, or
nearly so. It is evident from Figs. 254 and 255 that radial blades
willnot be moving edgewise with respect to still water at the time
they reach the water surface. This result may, of course, be ac-
complished by setting the blades at suitable fixed angles. But fixed

blades so set are usually regarded as undesirable, perhaps without

good reason.
In the United States the development of wheels which will not
suffer from excessive obliquity of blades at entering and leaving
has been toward wheels of large diameter and wide narrow floats
of small immersion. This line of development was facilitated by
the type of engine usually fitted on paddle steamers.
Furthermore, broadly speaking, paddle steamers in the United
States have been for service in smooth waters, and hence could be

designed for a small immersion of floats which would be inadvisable

in rough water service.
3. Feathering Blades. In Great Britain, influenced perhaps
originally by the fact that many of the finest and fastest paddle
steamers were for service across the English Channel and had to be
prepared to encounter rough weather, paddle wheels are almost uni-
versally fitted with feathering blades.
As indicated diagrammatically in Fig. 256, a blade is pivoted about
its center, the pivots being carried by the framing of the wheel

proper, which revolves about A. Each blade has an arm perpen-

dicular to it on its back, to which is attached a link, and the other

end of the link is connected to a center K eccentric from A. The

point Kvery simply determined. The positions of H, G and F

are obviously fixed by the positions desired for a blade entering the

water, leaving the water and at maximum submergence. Then K

is the center of the circle passing through H, G and F.
It is very common in practice to fit feathering paddle blades as
indicated in Fig. 256, where the planes of the entering and leaving
blades intersect the circle of blade centers vertically above the shaft.
Paddle wheels have been fitted where the blades remained vertical
throughout the revolution, but this is not done now.
It might seem very simple fromFigs. 254 to 255 to determine
the proper angles for blades entering and leaving the water, but the
actual problem is one of extreme complexity. Figs. 254 and 255
show velocities with reference to water at rest, and this is far from
the conditions of practical operation.
The water upon which a paddle wheel acts has been previously
disturbed by the ship, the amount of disturbance varying with the

speed. Moreover, each paddle enters water which has been dis-
turbed by the preceding paddles. There is little question that in
practicallyall cases of side paddle wheels the paddles enter water

which has already a sternward motion. Stream line action and the
action of preceding paddles will both give the water a sternward
motion, and even if the wheel is located at a wave crest as is
desirable the forward motion due to the wave motion will be less
than the other two.
For stern wheels stream line and wave action will give the water
a forward motion, the action of preceding paddles a rearward motion,
and it is not possible without extensive experiments to lay down any
general conclusions.
4. Comparison of Fixed and Feathering Blades. Paddle wheels
with feathering blades are heavier, more complicated and more
expensive than wheels of the same size with fixed blades. But in
practice they can be made materially smaller in diameter for the
same efficiency, and also can be given greater depth of immersion
resulting in a larger virtual area of
paddle for a given actual size.
This an important consideration for high-speed paddle vessels.

The smaller the wheel the higher the engine revolutions, and it is
usually desirable as regards weight and space to increase the revo-
lutions of paddle boat engineswhen directly connected. In practice
fast high-powered paddle boats are usually fitted with feathering

blades, fixed blades being used when the revolutions are low and
the diameter of wheel great, or for service in remote rivers where

simplicity is essential.

5. Paddle Wheel Location. While not proposed to con-

it is

sider structural details, some considerations affecting paddle wheel

design will now be taken up. In practice, paddle wheel vessels are
side wheelers or stern wheelers. In side wheelers the wheels are
located somewhere near the center of length. It is advisable to

locate them so that they work in a crest of the transverse waves

caused by the ship, or at any rate not in a hollow. When working
in a crest there is a virtual wave wake favoring a
efficiency, while in
hollow the wave wake is prejudicial to efficiency. The stream line

wake in which side wheels work is prejudicial to efficiency, so that

side paddle wheels usually have a virtual negative wake. Also the

wash from the wheels increases the frictional resistance of the rear
of the ship and produces a virtual thrust deduction.
Side wheels cannot be placed very far forward or aft of the center
of ships of ordinary form without danger of under or over immer-
sion through changes of trim, incident to service.
Stern wheel boats are of the wide flat type and the draft aft
does not vary much in service.
Stern wheels are so located that the wake due to stream line and
wave action is in their favor, and they will cause but little thrust
deduction as a broadly speaking, the stern wheel may
rule, so that,
be expected to be more efficient as an instrument of propulsion than
side wheels.
It is very desirable to fix the heights of all paddle wheels so that
the desired immersion will be had when the vessel is under way.
This can readily be done by model basin experiments in advance,
and for the best results with feathering wheels the question of blade
angles at entrance in and departure from the water should also be
investigated experimentally.
The immersion of paddles is varied somewhat with the service.
For seagoing boats the immersion of the upper edge of the paddle
in its lowest position is seldom less than \ its breadth and as great
as .8 its breadth. For smooth water service the immersion is usually

less, i to i the breadth. The desirable immersion depends some-

what upon the type of float. A very long narrow float on a large
wheel may have its upper edge immersed its whole breadth without
loss of efficiency.
6. Dimensions and Proportions of Paddle Wheels. One of the
most important questions arising in the design of any type of paddle
wheel is the determination of the dimensions of the blades, buckets
or floats, as they are variously designated.
These are sometimes curved, but seldom curved much, and may
be taken as rectangular. The length or horizontal dimension of
the float isalways greater than its width or radial dimension.
There is found in practice a difference in proportions between
feathering and fixed floats. For feathering floats the length is
usually about 3 times the width, though shorter floats have often
been fitted. For fixed floats the length is seldom less than 4 times,

and may be in extreme cases 7 or 8 times the width. This difference

of practice naturally arises from the fact that floats are usually
made as long as possible from practical considerations, as tending to

efficiency, and then as wide as necessary to absorb the power. For

side wheels, floats are, however, seldom longer than the beam
even for vessels always in smooth water, and for seagoing vessels
it is not regarded as good practice to make them longer than about

f the beam.
The float area is dependent primarily upon the power absorbed
and the slip. We have seen that the theoretical formula involved
P = I
K'. This may be rewritten A = K where A is area of
two on each side) in square feet, 7 is indicated horse-
floats (one

power and proportional in a given case to P, V is speed of ship in

knots and K is a coefficient depending primarily upon the slip and
secondarily upon a large number of minor factors, such as wake,
thrust deduction, float proportions, number and immersion, etc.
Hence K may be expected to vary a great deal from ship to
ship, but fortunately it is not necessary to know it with minute
Analysis of a number of published trial results for paddle steamers,
nearly all with feathering floats, appears to indicate that a reason-
able expression for the average value of K
will be, for slips used in

practice ranging say from .10 to .30,

K = 212.5
~ 375 *

From the nature of the case individual values of K may be

expected to vary materially from the average. A long narrow blade
deeply immersed may be expected to show a much smaller value
of K
than a short wide blade with its upper edge barely immersed.
Then a suitable paddle area may be determined approximately

by the formula .4 = (212. 5 3755) It must be remembered

that in the above A is total area in square feet of two paddles when
side wheels are fitted, and 5 is slip based upon the peripheral velocity
of the centers of paddles.
It is desirable to keep the slips of paddle wheels low. For feather-

ing floats .15 is frequently aimed at, and for fixed floats .20. Know-
ing the speed of the ship and the desired slip, the peripheral velocity
of the mean diameter of the paddle wheel upon which slip is based
is known, and this in conjunction with the desired engine revolutions
fixes the mean diameter of the wheel.
desired float area being known, the float dimensions are

determined, enabling all dimensions to the wheel to be fixed. If

these are found suitable the desired blade angles at entry and depar-
ture will govern the details of gear for feathering blades when such
are fitted.
As regards number of blades it is a very common practice with
fixed blades to fit one for each foot of outside diameter of wheel.
This number should not be exceeded for wheels of good size and may
be reduced by 20 per cent or so without detriment. The spacing
of feathering blades is greater than that of fixed blades, partly be-
cause such blades are usually relatively deeper than fixed blades
and partly because of the additional complications of feathering
gear for blades close together.
With feathering blades there are sometimes fitted one for each
foot of radius but a greater number
are usually regarded as desirable,

say about 3 blades to each 2 feet of radius.

33. Jet Propulsion

i. General Considerations. Jet propulsion has never been used

except experimentally. In jet propulsion water is taken into a ship,
where passes through some form of pump or impelling apparatus

and then delivered astern through suitable pipes. Many schemes

for jet propulsionhave been brought forward in the past, usually
including methods for diverting the jets sidewise as desired, in order
to gain maneuvering power.
While some schemes of jet propulsion have been actually tried,
none has proved so efficient as the screw propeller or paddle wheel.
Hence, jet propulsion is of academic interest only and will not be
given detailed consideration.
That any system of jet propulsion involving any form of impelling

apparatus known at present must be inefficient will be evident from

Fig. 171. It will be found from this that, even with frictionless

impelling apparatus, if there is not to be a great loss through slip

the pipes to get the water into and out of the ship must be so large
that they will involve very serious increase in skin friction to say
nothing of eddy losses. If pipes are made small there is unavoid-
ably a great loss by slip, and still larger loss by friction in the pipes.
Furthermore, any pump or impelling apparatus now known is not
materially more efficient in communicating velocity to a given

quantity of water than the screw propeller or the paddle wheel.

Hence, jet propulsion, involving taking water in large amount into
the ship and discharging it again, is with any known form of impell-

ing apparatus necessarily less efficient than the screw and the paddle,
which operate in the water outside the ship.
Since the essential inefficiency of jet propulsion as a method of

utilizing the power of ordinary engines has become evident, some

inventors have attempted to devise apparatus specially adapted to
jet propulsion in which power is developed more economically than
in engines driving propellers and paddle wheels. Efforts along this
line have not hitherto been successful.


34. Measured Courses

i. Features Desirable for Measured Miles. Trials for the

determination of speed must be made over a course of known length,
unless by trials already made over such a course the relation between
revolutions of the propellers and speed through the water has been
established so that a speed trial may be conducted in free route.
The measured course may be long or short. The difficulties of locat-

ing, measuring and marking a satisfactory long course are evidently

much greater than for a short course, and nearly all accurately
measured and marked courses are one nautical mile long. For a
number of years, however, four-hour full-speed trials of United
States naval vessels were held on long deep water courses extending
to the northward of Cape Ann on the Massachusetts coast. The
length used was carefully determined in each case so that the vessel
would run about two hours in each direction and four or five vessels
or more were anchored on the course for the double purpose of
defining it and of making observations of the tidal current during
trials. Of late years,however, four-hour full-speed trials have been
made in free route by the standardized screw method. For stand-
ardizing the screw or determining the relation between speed and
revolutions, trials are usually held on a course one measured mile
in length near Rockland, Me. This course is shown in Fig. 257.
It is seen that the course is defined by four range buoys, one at
each end of the measured mile and one a mile from each end. These
buoys, however, are for steering purposes only. The ends of the
course are fixed by ranges established on shore, each with a front
and rear signal or beacon. When these signals are in line the
observer is at one end of the course, which, as shown, is perpendicular
to the range lines.

The desirable features for a measured mile course in tidal waters

are enumerated below.
If they were all present in any particular case the course would
be ideal. In practice it is necessary to be satisfied with a reasonable
approximation to the ideal.
1. The range marks on shore at each end of the course should
be well separated say f the length of the course or more
and should by the transit of the front signal past the back signal
mark definitely and sharply the instant of crossing the range. This
is best attained when both front and back signals show against the
2. The situation should be such that the course is not far from
shore and fairly well protected, insuring smooth water when the
local wind conditions are favorable.

3. There should be plenty of room at each end of the course for

4. The course should be so situated that the ship making runs
over it need never cross or obstruct a channel or fairway that is

much used.

5. The tidal current should be small and always parallel to the

6. The depth of water should be sufficient, so that the resistance
of the ship using the course is practically the same as in deep
As regards most of the features enumerated above, the Rockland
course, shown in Fig. 257, approximates fairly closely to the ideal.
It has the disadvantage of being rather remote from most of the

building yards whose vessels must use it.

It would be better if the front and back signals marking the ranges
were further separated and showed above the sky line. It may be
noted in this connection that if the range marks do not show against
the sky a course running north and south is not so good as one

running east and west. If the ranges are to the west of the course
the marks are difficult to pick up in the afternoon, and if they are
to the east they are difficult to pick up in the forenoon.

35. Conduct of Speed and Power Trials

i. General Considerations. Vessels may be given many kinds
of trials, as of speed and power, of fuel economy, maneuvering capa-
city, etc. We need consider the first named only.
Speed and power trialsmay be considered from the point of view
of (a) the owner, (b) the designer, or (c) the builder. In some cases,
as for vessels of war government establishments, the owner,
built in

designer and builder are one; frequently for vessels of war the owner
and designer are one; and usually for merchant ships, and sometimes
for vessels of war, the designer and builder are one.
From whatever point ofview we consider speed trials, however,
they are primarily of importance for new and untried vessels. For
such vessels the owner wishes to know what his ship will do in
service and from the results of progressive speed and power trials

he can generally closely estimate the results to be expected in ser-

vice. The designer wishes to know what the ship actually does
under known trial conditions in order that he may utilize the infor-

mation in preparing subsequent designs. The builder is generally

required to guarantee certain results to be demonstrated by trial
before the ship leaves his hands and at times wishes to develop on
trial certain results not exacted by his contract, but which may be
of use to him in a business way. Apart from this he is apt to con-
sider that trials conducted at his expense should be reduced to the
lowest terms.
As a result of various conflicting considerations the most that
can usually be expected for speed and power trials of a new ship in
the builder's hands is the determination of corresponding values of
speed, revolutions, and power over a reasonable range from the
maximum down, at one displacement and under favorable condi-
tions of wind and weather. Such a trial is usually called a pro-
gressive speed trial and appears to have been first developed in
Great Britain by Mr. William Denny. Concerning this develop-
ment Mr. William Froude said in a paper before the Institution of
Naval Architects, April 7, 1876:
Mr. Denny has taken the bold but well-considered step of dis-
carding the conventional type of measured mile trials which, as

regards the speeds tried, have long been limited to full speed and
half boiler power. Mr. Denny now tries each of his ships at four
or even at five speeds; and the result is that he obtains fair data
for a complete curve of indicated horse-power from the lowest to
the highest speeds; whereas with trials on the ordinary system we
obtain merely two spots in the curve, and these at comparatively
high speeds, the intermediate or lower portion of the curve being
2.Accuracy Possible in Progressive Trial Results. The deter-
mination of accurate results on a progressive trial is by no means
the simple matter it might seem at first. Approximate results are,
of course, readily obtained, but for the results of progressive trials
to be of real value for the designer they should be quite accurate.
What we need are simultaneous values of speed of the vessel, power
indicated by the machinery and revolutions per minute of the en-

gines, determined for a sufficient number of speeds covering a good

range to enable accurate curves of power and revolutions as ordi-
nates to be drawn on speeds as abscissae throughout the range
covered by the trials.
If we had available a measured course in perfectly still, calm, deep

water, and wished to determine the most reliable curves from a defi-
nite number would evidently be desirable to run back
of runs, it

and forth, increasing and decreasing the speed or revolutions by

equal amounts between successive runs. Observing on each run
the time and revolutions on the course and taking indicator cards
for the power determination, we could plot curves through points
obtained by the observations.
Progressive trials are not made on ideal courses, as above. Even
if they were, it would seldom happen that the data obtained would

be absolutely consistent and harmonious. It is probable that on a

course in still water the time on the course would be determined
with a good deal of accuracy. But even with a long straight run
at each end before coming on the course an important point fre-
quently neglected the speed on the course is seldom absolutely

uniform. Unless steam actually blowing off all the time the boiler

pressure is always going up or down it may be very slowly with

skilled firing, it may be with sufficient rapidity to cause quite an


appreciable change in speed while on the course. Moreover, the

rudder constantly being used more or less, and even when put

over to a small angle only it has a noticeable effect upon the speed.
This is a matter of practical importance in the conduct of trials
which does not always receive proper attention.
Then the indicator even the best is not an instrument of

precision. If several sets of cards are taken during a run the powers
worked out from them will differ materially. Professor Peabody,
an authority on indicators, considers that even under favorable
circumstances the unavoidable error of a steam engine indicator
is likely to be from two to five per cent."

If the indicated horse-power is determined on the measured course,

not less than three sets of cards should be obtained and the average
of all good cards used in determining the average power. At times
some cards are obviously defective, and these should be thrown out.
For single-screw ships the revolutions and speed vary together,
and there are no serious complications from the inevitable slight
variations in revolutions, except that sometimes there is doubt as
to the proper revolutions to use with the indicator cards for the
determination of power. But with twin-screw ships the revolutions
and power of the two engines are not identical on any run. The
only thing that can be done in such cases is to try to have the port
and starboard revolutions during each run as nearly the same as
possible and use the average of the two results. With two screws,
unless the propellers differ more than they should, we may safely
assume that at a given speed and the same revolutions, each engine
will require the same power. In practice, owing to minor differences
in propellers, and differences in engine friction, the assumption is
not exact. But it is near enough, and is, one we
in fact, the only
can make.
With three screws, however, the case is different. At full speed,
with everything wide open, the central engine will differ from the side
engines as regards both power and revolutions, even if identical in
size with them. When itcomes to the runs at reduced speed, we
may for a given speed have enormous variations in the power dis-
tribution. It would seem proper in such cases, where the engines
are identical, to be careful to have the steam pressure in the H.P.

valve chests and the linking up the same for all three engines on
each run. Otherwise the curves of slip of the center and side
screws will be very erratic. With four screws the case is even more
For such vessels, where each engine is independent, it may be
necessary to plot results upon speed plotting separate curves
of revolutions for each engine. But even here equally good results
can be obtained by plotting results upon the average revolutions
of one pair of engines plotting, upon these revolutions, a curve
of the average revolutions of the other engine or pair of engines.
For turbine installations, where the turbines are in tandem, the
steam passing from one turbine to another, this method is distinctly
When we come to turbines we meet the difficulty of determining
the actual power exerted by them. Several methods are used
all based upon the fact that the twist of the shafting is proportional

to the torque of the turbine. This twist is a small quantity in any

case, and accurate determination experimentally is difficult. It

isprobable, however, that as the use of turbines extends the accu-

racy of their power determination will be improved. With an
accurate torsion meter the determination of shaft horse-power will
be much simpler and easier than the determination of indicated
horse-power by means of indicators.
3. Elimination of Tidal Current Effects. It is evident from
what has been on an imaginary still-water course a
said that even

progressive would
trial not be free from doubts and difficulties in

connection with obtaining and plotting the results.

Actual measured courses, however, must be laid off in a tideway
where tidal currents varying in direction and magnitude are encoun-
tered. No suitable for a progressive trial unless the tidal
course is

current practically parallel to the course.

is Slight cross currents
are nearly always present, however. When they are present the
steering on the course should always be by compass and not by
buoys or other fixed fore and aft ranges. By always steering a
compass course parallel to the true range the effect of slight cross
currents is eliminated. So we will consider from now on only the
current parallel to the course. Suppose, first, that the current is

constant and that we make two runs at the same true speed one
with and one against the current.
Suppose V is the true constant speed of the two runs, C the con-
stant but unknown speed of current and V\, F 2 the apparent speeds

of the successive runs. Then Vi = V + C, V Z =V C whence V =

% ( V\ +V
z) or the true speed through the water is the average of the

two apparent speeds with and against the current. Sometimes the
true speed is taken as that corresponding to the average time of
the runs with and against the current. This is incorrect. The true
speed for two runs with and against a constant current being the
average of the two apparent speeds, it is a common practice to
make the runs of a progressive trial in pairs one run being made
in each direction at the same speed. There are two objections to
this. One is that the tidal current changes between runs. The
other often more serious in practice arises from the fact that
in practice the successive runs are made not at the same speed but
at different speeds, and the average horse-power is not the proper

horse-power for the average speed. Figure 260 illustrates this, in

an exaggerated form. A and B are points on a curve of horse-
power plotted on speed corresponding to two runs. C is the point
on the curve corresponding to the average speed, while D, midway
of the straight line joining A and B, is the average horse-power.
The firstsource of error, the change of tidal current, can be largely,
but not entirely, eliminated by making a series of runs over the
course at one speed and obtaining the true speed from the apparent
speeds by the method of successive means. This is illustrated below
with four runs the apparent speeds being Vi, Vz V 3 F
, , 4.


of first means, and so on. There appears to be a difference of

opinion as to whether, when there are more than four runs, the true
speed should be taken as the final mean, or the average of the
second means. As appears above, for four runs the two are the
Now if n denote the number of a run of a series we can always
express C, the strength of the current, in the form
C = a + bn + cn*+ dns + en +
. .

using as many terms as there are runs in the series. Suppose, for
instance, there are four runs. Then we have
C = a + bn + cnz + dn3 .

Denote by C\, C 2 C 3 C4
, ,
the actual current strength of the four
successive runs.

Then d= a + + c + d, b
C2 = + a + 4c + 8
2 b </,

C3 = a + 36 + gc + 27 d,
C4 = a + 4 b + 16 c + 64 d.

These are four equations from which we could determine the four
unknown quantities a, b, c, d. Hence, no matter what the current
strength of the successive runs, we could always find values of the
coefficients a, b, c and d such that we can represent the current by

C = a + bn + cn z + dn3 .

On solving the equations above for a, b, c, d we have

a = H 2 4 Ci- 36 C 2 + 24 C 3 - 6 C4 ),

b = $ ( - 26 Ci+ 57 C 2 42 C 3 + ii C4 ),
c = i(9 C"i- 2 4 C2 + 21 C 3 - 6 C4 ),
<* = H-C" 4- 3 C 2 -3C 3 + C4 ).

Now consider further the final mean result. We have, if V

denotes the true constant speed of the four runs,

F 1= F + C 1= F + a+ b+ c+ d,
V2 = V - C 2 = V -a - 2b - 40- Sd,
V3 = V + C 3 = V + a + sb+ 9 c + 2 7 d,

F4 = V - C4 = V - a - 4b - i6c - 64 d.

Final mean = \ (V\+ 3 F2 + 3 F 3 +F4 ). Upon substituting for

Vi, Vz, etc., in this expression, their values above in terms of V
and the coefficients a, b, c and d, we finally have, after reduction,

Final mean = V -ld=V -\(- Ci+ 3 C2 - 3 C3 + C 4 ).

In case only three runs are made the current formula is

C = a + bn + cn~,

and the currents of the successive runs are Ci, C2 and C 3 . For this
Final mean =i (Vi+ 2F 2 + F 3 )= F + Jc = V + i(Ci- 2 C2 +C 3 ).

Then the final mean

not the true speed unless the rate of change

of the tidal current and the timing of the runs is such that for four
- (d-C ) + 3(C 4 2 -C 3)
= o,

and for three runs

Ci+ C 3 = 2 C2 .

This will happen exactly only by accident. Another way of express-

ing the condition is that d, the coefficient of w4 should be = o, the ,

actual error being 1 d. As a matter of fact, in most practical cases

d would be very small and the final mean but little in error if the
assumptions upon which the final mean method is based were cor-

These underlying assumptions are two, namely, that the tidal

current varies according to a fair curve and that all runs back and
forth are made at the same speed.

Every one who has often plotted results of speed trials in a tide-
way will have encountered results which could be explained only on
the theory that the tidal current varied by fits and starts rather than

according to a fair curve.

It is sometimes assumed that the tidal current varies from maxi-

mum to minimum in a manner such that a curve of tidal strength

plotted on a base of time would be a curve of sines. This is per-

haps a reasonable approximation to the general outline of the curve,
but observations of actual strengths of tidal currents appear to show
that they vary erratically and would seldom plot as a fair curve
closely approximating a mathematical curve of sines.

A more serious error than that due to tidal current is liable to

result from the fact that successive runs of a group are not made
at the same speed. This is a matter of practical experience. It is very

unusual, indeed, for four successive runs to be made over a measured

course where the revolutions per minute, if accurately determined,
do not vary appreciably. If the speed were constant, the revolu-
tions should not change. Suppose now four successive runs were
made aiming at a uniform speed of ten knots, while the actual
speeds were 9.72, 10.24, 10.16, 9.88. The true average speed would
be ten knots, but the final mean of the four speeds above would be
10.1 knots. This is quite a large error. In the above I have not
taken account of the tide. The error is not affected if the tide is

such that the final mean would eliminate the tidal error if the runs
were made at constant speed. For instance, suppose the tidal cur-
rents were in knots .61, .74, .89, 1.06. For ten knots true speed
the apparent speeds would be 10.61, 9.26, 10.89, 8.94. The final
mean of these four speeds is 10 knots, as it should be. But if the
true speeds of the successive runs were as given above, the apparent

speeds after making allowance for currents, would be 10.33, 9-5>

11.05, 8.82. The final mean of these is 10.1 knots, as before.

Evidently, then, as the final mean method is equivalent to giving

the two middle runs of a set of four a weight of three as compared
with a weight of one for the first and last runs, when it is used for
speed it should in theory be also used for revolutions and power.
Thus, if a middle run of a series of four is made at a true speed
above the average the excess speed in determining the average speed
is given a weight of 3. This run will show excess power and revolu-
tions, and the average power and revolutions are properly to

correspond with the average speed by the final mean method the

power and revolutions should be given the same weight as the speed
in determining the average. Practice in this respect appears to be
somewhat variable. We often, but not always, find the final mean
method used for revolutions. It appears to be seldom used for
4. Methods of Conducting Progressive Trials. We seem war-
ranted in concluding that when we attempt to get a spot on a speed
and power curve by applying the final mean method to the data

observed during a series of four runs, we by no means eliminate

the probabilities of error. The question arises whether there are
not better methods, or simpler methods equally good. We wish to
determine curves as accurate as possible expressing the simul-
taneous values of speed, revolutions and horse-power. Now in any
particular casewe can usually determine the revolutions with great
accuracy. We can determine the indicated horse-power with reason-
able approximation, and with good indicators the error is as likely to
be in excess as in defect. For twin-screw vessels,when the two en-
ginesshow different revolutions during a run, the best we can do is to
take the total indicated horse-power as corresponding to the average
revolutions of the two engines. For any run we can determine the
speed over the ground with ample accuracy, but owing to tidal cur-
rent we cannot determine accurately the speed through the water.
Now in plotting our results shall we plot power and speed on revolu-
tions, or power and revolutions on speed, or perhaps speed and revo-
lutions on power? A little consideration will show that there are
real advantages in using revolutions as the independent variable,
so to speak, and from the trial data plotting on revolutions separate
curves of power and speed. For the revolutions of a run can and
should be determined exactly to all intents and purposes.
Then by plotting our approximate data upon the correct revolu-
tions we get rid of one element of uncertainty. We do not, for
instance, plot a spot for power where the error is in excess over a
spot for speed where the error is in defect. We will ultimately

arrive at a more reliable relation between speed and power by deter-

mining first the most reliable relation between each and the accu-

rately determined revolutions. Starting, then, with the basic idea

that we will in the first place plot speed and power as ordinates
upon revolutions as abscissae, how should the progressive trial be
conducted in order to determine most reliably the relation between
power and revolutions ?
We know from the Theory of Probabilities that if we wish to
determine a single quantity as, for instance, the value of a fixed
angle the best plan is to take as many observations as possible
and use the average as the best obtainable approximation to the
true value. Similarly, if we wish to determine a curve from experi-

ment the best plan is to ascertain as many approximate spots as

possible, plot them and draw the final curve as the average curve
through the spots. Then to establish a curve of power on revolu-
tions we should make numerous simultaneous determinations of

power and revolutions, plot the results and draw an average curve
through. To locate the curve of power as accurately as possible
from a given number of runs, it would be better to have each run
made at different revolutions. This would enable us to cover the
curve closely with experimental spots. Here we encounter another
weak point of the four-run final mean method.
Sixteen runs are necessary to determine four spots on a power
curve, and four spots are insufficient for the accurate determination
of a curve of power covering a wide range of speed. On the other

hand, sixteen spots distributed at approximately equal intervals

over the whole length of the curve will locate it with great accuracy.
Each spot may be in error, owing to limitations on accuracy of any
determination of indicated horse-power, but if the errors are as
likely to be positive as negative a fair average curve through sixteen
spots will practically eliminate the indicator errors. If the indi-

cators have a constant positive or negative error no number of

experimental spots will eliminate it. I conclude, then, that as to

the relation between power and revolutions about sixteen simulta-
neous determinations of revolutions and indicated horse-power,
made at approximately equal intervals of revolutions, will enable a

satisfactory power revolution curve to be drawn. These observa-

measured course, when the
tions need not necessarily be taken on the

speed revolutions observations are being made. It is usual, how-

ever, to take the indicator cards while on the course. When the
observing staff is adequate it is more convenient to make one job
of it, and if the water on the measured course is somewhat shallow,

so as to affect the results, it is desirable to determine everything

under the same conditions. By doing this, too, we avoid the chance
of the initial friction of the engines altering between two sets of

runs, one to determine the power revolution relation, and the other
the speed revolution relation. Finally, with an ample observing
staff the time of a run over the measured course is generally of a
length convenient for taking several sets of cards. There is, how-

ever, something to be said in favor of making runs off the course

for determining power revolutions spots. With a small observing
staff indicator cards can be taken more at leisure and given revolu-
tions can be maintained until a sufficient number of satisfactory
cards are taken, even indicator accidents crop up.
Again, as
soon as good cards for a given number of revolutions are obtained
the revolutions can be changed at once up or down. This will
not save much
time at high speeds, but will at low speeds, so that the
total time the staff must be kept at the indicators will be a good

deal shorter. The preferable method really seems to depend in the

end upon the observing staff available. With an ample staff of

skilled observers, so that in addition to time and revolutions on the

course three good cards can (barring accident) be obtained during
each run from each end of each cylinder, it would seem advisable
to make all observations on the measured course. With a small
staff of observers, however, including many without good experience
in such work, it would often be advisable to run separate trials,
making the progressive power revolution trial before or after the
speed revolution trial on the measured course.
Fig. 258 shows trial spots and final curve of power on revolu-
tions as drawn from the trial of an armored cruiser.
Let us consider now the most suitable practical method of deter-
mining the speed-revolution relation from trials on the course. In
the first place, no method will give reliable results unless we have a
sufficient number of runs. Each experimental spot is necessarily
and unavoidably somewhat in error. Hence, in order to get a
reliable curve we must have so many spots and have them so close

together that the accidental and erratic errors are practically elimi-
nated by drawing a mean fair curve. There are two methods which
may be used with confidence. The first is probably the most accu-
rate and reliable, provided the trial is conducted with special skill

along the lines described below. It is also adapted to the determina-

tion of the power revolution relation by the method just given. The
second method is probably preferable -for the usual run of trials.

Under the method make a series of runs back and forth


alternately with and against the tide and increasing or decreasing

the revolutions by equal increments after each run. The curve of

true speed then will fall midway between the two curves of apparent

speed, one with and one against the tide. The advantages of this
method are that if the curve of tidal variation is a fair curve and the
trial skillfully run so that the interval between successive runs varies
according to a fair curve, all spots of apparent speed will fall upon
fair curves. Should, however, a spot be erratic, it will naturally
fall off the curve and be given little weight in drawing the final

curve of apparent speed. It is a very real advantage in such work

to have a method of reducing the data such that bad spots show for
themselves and are not incorporated in the final results. It is evi-
dent, however, that to get reliable curves of apparent speed we
should have a sufficient number of spots for each curve. Not less
than sixteen runs in all should be made. Figure 259 shows curves
of apparent speed with and against the tide and the mean curve
from the trial of an armored cruiser. All experimental spots are
There are some objections to the above method. One is, that at
top speed, the most important part of the curve, we would have
only one run, and the high speed part of the curve would not be
defined so well as the lower portion. This difficulty should be over-
come by making three runs at top speed two in one direction, and
one in the other, and determining the final speed of the three by
giving the middle run double the weight of the others. This is
equivalent to taking the second mean of the three runs. The other
objection to this method is that for thoroughly satisfactory results
a once begun should be completely carried through without

stopping. This sometimes introduces practical difficulties. A run

may be lost through breakdown of the observing apparatus or inter-
ference of some other vessel while on the course. This is not a
very serious objection, because it is found in practice that even if
the intervals between the runs are somewhat erratic the curves of
apparent speed can be drawn tolerably well. Another objection of
the same nature is that if a trial is interrupted after five or six runs
the results of these runs are of little value, as they are not suffi-
ciently numerous accurately to define the part of the curve to which
they refer, and a whole new trial has to be made.
The second recommended method of running a progressive trial

is to make runs in groups of three, 18 in all for a fast vessel, 15

for a vessel of moderate speed and 12 for a slow vessel. Each group
should be made at a constant number of revolutions, as nearly as

possible, the revolutions for the various groups covering the range
desired. Then, taking for each group of three runs the second mean
of speed and revolutions we have for a fast vessel six spots through
which to plot a curve of speed and revolutions. This method in
practice gives from each group of three runs a spot substantially as
reliable as if four runs had been made. While it has the advantage,
as compared with the four-run method, of giving more spots on the
curve for a given total number of runs, it also has the advantage of
beginning consecutive groups with runs in opposite directions.
That is to say, if one group began with a run to the north the next
group will begin with a run to the south. This is a desirable con-
dition, as tending to eliminate some of the errors due to tidal current.
This method has the advantage of requiring less skill and care in
the conduct of trials, and each group of three runs stands by itself

and is not wasted in case it is necessary to stop the trial. It is not

quite so accurate as the method previously described, but the dif-
ference in accuracy would not be appreciable in the majority of
cases. A practical advantage is that it does not require readjust-
ment of throttles and links after each run in order to change the
revolutions. This adjustment, in order rapidly to change revolu-
tions by a definite amount, is by no means the simple matter it

might appear at first thought and requires quick and accurate work
in the engine room.
If there were no variations of tidal current between runs both

methods above described would be theoretically exact. It is evi-

dently desirable to time the progressive trial so that during it there
shall be as little variation of tidal current between runs as possible.

Now, when the tidal current is at a maximum, whether ebb or flow,

the variation of current is at a minimum, while about the turn of
the tide the rate of variation is about at a maximum. This state-
ment would be the curve of tidal current plotted on
exactly true if

time was a curve of sines, as often assumed, and is substantially

correct even as applied to actual tidal currents, varying by leaps
and bounds rather than with definite progression. Then a pro-

gressive trial should always be run during the strength of one tide.
A trial can generally be run in four hours or less, and so should, if

practicable, be begun about an hour and a half after the turn of the
tide. Circumstances often render this inconvenient or impossible,
and weather conditions frequently cause the turn of the tide to
come before or after the time fixed by tide tables, but the best time
for a trial should be used unless there are good reasons to the con-

So far as accuracy of results is concerned it makes no difference
whether we begin with the low speeds and work up or with the high
speeds and work down. It seems advisable, however, as a rule to
begin with the top speeds and work down. With clean fires and
fresh men the top speed can be obtained and maintained with more
ease than after several hours of running. Also, if the trial is spoilt

by a breakdown it is more apt to come during the high speed runs,

and if a breakdown must come it is better to have it come early
than late.

There may be mentioned here some minor points in connection

with the conduct of trials which tend to produce accurate and sat-

isfactory results. It is desirable after a run to shift revolutions

promptly to the revolutions for the next run, if they are to be

different. a pressure gauge giving the pressure in the
If there is

H.P. chest (beyond the throttle) it is easy by preliminary runs to

establish a curve (or curves, if more than one valve gear setting is
to be used) giving the relation between H.P. valve chest pressure
and revolutions. Then it is necessary only to establish the proper
pressure to insure that the revolutions are sufficiently near what
is desired. Such a pressure gauge as above is apt to fluctuate vio-
lently unless its cock is nearly closed. Systematic handling of the
ship when off the course is desirable. Each time when coming on
the measured course the ship should have made a long straight run
with the minumum operation of helm. For most trials about a
mile a convenient and desirable length for the straight run, and

it much facilitates trials if in addition to buoys at the ends of the

measured course, moored closely on the ranges, there are planted
buoys in the line of the course a mile from each end. Suppose we
have the course thus buoyed as indicated in Fig. 257. Before begin-

ning the trial proper while warming up steam over the course
as indicated in Fig. 257 by ABCDEFCBGHA.
When abreast the buoy D put the helm over to a moderate and
definite angle, say 10 degrees. Steady the ship on the course EF
which will cross the line of the course a little beyond the buoy C.
While on this course note carefully the compass reading and deter-
mine the reading of the steering compass which will give the opposite
course FE. Then when coming of! the course at C after a run, put
the helm over at once and steady the ship on the course FE. If

the revolutions are to be changed for the next run the engine room
force should immediately set to work on this. With skillful han-
dling the new desired revolutions should be attained before the vessel
is at E. If this is so, on reaching E abreast the buoy put the D
helm over to 10 degrees. The by the time she swings
vessel will,
to the correct heading for the next run, be practically on the line
of the course, requiring very little use of the helm to come dead on.
If the revolutions are not adjusted by the time the vessel reaches
E, she should at this point be steadied on the course EK, shown
dotted in Fig. 257, and kept on this course until the revolutions are
satisfactorily adjusted or the vessel has run so far that there will
be ample time after turning finally to adjust the revolutions before
the vessel reaches D. The methods are of course just the same at
each end of the course.
To conduct a trial in this way requires quick communication and
complete understanding between the deck and the engine room, but
results will be distinctly superior to those obtained by more hap-
hazard methods.
5. Trial Conditions. It is customary to make progressive trials

with clean bottoms under good conditions of wind and sea. For
men-of-war the trial is
generally made
at normal load displacement.
For merchant vessels the displacement is ^sometimes the average

displacement to be expected in service, but generally a less displace-

ment and at times a very light displacement.
The usual practice is at times criticised. As to men-of-war, for
instance, alleged that they will never show in service such good
it is

results as upon trial. It is true that there is ever present the temp-
tation to run trials at too light a displacement. This is largely due

to the natural desire of those concerned to make the best showing

possible. But the loss of speed in service due to increased displac-
ment is apt to be exaggerated, particularly for large ships. More
potent causes are rough water at sea, dirty bottoms, poor coal, or
inability of the engineering personnel to get good power results. It
is evidently desirable to have trials always run under uniform or
standard conditions. The most easily attained standard trial con-
ditions are obviously fair weather, smooth water and a clean bottom.
From reliable results under such conditions the results which should
be attained in service can be estimated with sufficient approxima-
tion until they can be ascertained experience. As a general
thing, however, progressive trials cannot, and are not expected to,
show exactly what a ship will do in service. This requires service
experience. They furnish data to enable the performance of the
ship under standard conditions to be determined and compared with
other vessels, and in case the performance is poor careful progressive
trials will not only determine that fact, but as a rule, upon analysis,

indicate the line that should be followed to obtain improvement.

36. Analysis of Trial Results

i. Components of Indicated Horse Power. Figure 260 shows a

curve of speed and power for the U. S. S. Yorktown, the powers as
ordinates being plotted over the speeds as abscissae.
The power is the indicated horse-power developed in the cylinders
of the engines. We know that only a fraction of this power is finally
utilized to propel the ship and it is important to gain some idea of
the distribution of the remainder.
The engine itself absorbs a certain amount of power through its
own friction. This friction is usually classed under two heads,
" " " "
namely, initial or dead friction, due to tightness of pistons,
" "
valves, glands, bearings, etc., and load friction, or the friction
due to the load upon the bearings and thrust block.
The power required to work feed, air, circulating and bilge pumps,
driven from the main engines, is usually classed with the initial
friction. For reciprocating engines, the power P delivered to the

propeller the original indicated horse-power less the power as


above absorbed by friction. For turbine engines the power is


usually determined from the twist of the shaft, measurements being

taken astern of the thrust block. All of this shaft horse-power is
delivered to the screw except what is wasted in friction of line bear-

ings, stern tubes, and outward bearings, if any. This is usually so

small that the shaft horse-power is assumed to be the same as the
propeller power P.
Of the propeller power P
a portion is wasted in friction and slip
of the propeller. The remainder is used in developing thrust horse-
power. Also there is added here a certain amount power derived
from the wake which also appears as thrust horse-power. Of the
thrust horse-power a certain amount is used to overcome the aug-
mentation of resistance of the ship due to the suction of the pro-
peller, and the remainder is the effective horse-power, the net

power required to drive the ship.

The above components of the I.H.P. vary widely. The initial
friction will absorb from as low as 3 or 4 per cent of the power in

large well-adjusted engines with independent air and circulating

pumps to 10 per cent or more in the case of machinery badly
adjusted with air and circulating pumps driven off the main engines.
The load friction is usually taken as about 7 per cent of the
remainder obtained by deducting the inital friction power from the
original I.H.P. With well-lubricated engines it is generally some-
what less. Investigations of the shaft horse-power of reciprocating
engines by means of torsion meters have shown as much as 92 per
cent of the indicated horse-power delivered to the shaft, involving
a loss of but 8 per cent for both initial and load friction. Engines
seldom run any length of time with excessive load friction. It

promptly causes hot bearings.

The ultimate distribution of the propeller power the shaft

horse-power for turbine jobs a question of the efficiency of the


propeller, the wake factor and the thrust deduction.

It is evident from what has gone before that as a reasonable work-
ing approximation we may assume that for a reciprocating engine
of high-class workmanship about 90 per cent of the indicated horse-

power delivered to the propeller when independent air and circu-


lating pumps are fitted, and about 85 per cent of the indicated power
when all pumps are driven off the main engine.

Accurate trial results can be analyzed to give an approximation

to the resistance of the ship, and hence efficiency of propulsion, etc.,
but these quantities can be estimated directly with sufficient accu-
racy and with much less labor by methods already given. It is
very desirable, however, to determine accurately the initial friction
of an engine, as then we know with close approximation the pro-
peller power, P, and this power is an essential factor of the propeller
design. Hence we will now consider in detail the initial friction of
an engine and methods for determining it from progressive trial
2. Initial Friction Determined by Curves Extended to Origin.
Mr. William Froude, the pioneer investigator of this question, defines
initial friction as the friction due to the dead weight of the work-

ing parts, piston packings, and the like, which constitute the initial
or low speed friction of the engine." The initial friction, or internal
resistance, is generally regarded as constant throughout the range of

speed and power of the engine, thus differing from the load friction,
which is generally regarded as absorbing a uniform fraction of the
power developed. As a matter of fact, it seems altogether probable
that the internal resistance varies slightly with power and revolu-
tions, but the variation is probably so small as long as bearings run
cool that we are justified in ignoring it.
There is no doubt that the internal friction will alter materially
from time to time, due to changes in tightness of various parts.
The problem under consideration, however, is the determination of
the initial friction at a given time. If the frictional resistance is
constant the power absorbed by it will be proportional to the revo-
lutions, so that if we denote by // the horse-power absorbed by
initial friction and R denotes revolutions, we have //= R X (a coeffi-

cient), where the coefficient at a given time for a given engine is

constant. Suppose we denote the coefficientby C/, then //= Cf R.
Now analysis and consideration of the various absorbents or com-
ponents of the indicated horse-power, such as the power utilized to
propel the ship, the power wasted by the propeller, the power
absorbed in ^load friction, etc., show that they except //, must

vary as some power of the revolutions greater than unity. This

being the case, it follows that if / denote the indicated horse-power

at revolutions R, we may write / = C/R + <

(/?), where we know
that (R)(f>
is some function which varies always
of the revolutions
as a power of R greater than unity. If, then, we plot a curve of 7
on revolutions, as we approach the origin the curve of 7 will ap-
proach the straight line If
= Cf R, and at the origin will be tangent
to this line. Hence C/ can be determined from the inclination at
the origin of the curve of 7 plotted on R.

Figure 261 shows for the U. S. S. Yorktown a curve of indicated

horse-power plotted on revolutions, the curve being extended to the
origin and the tangent at the origin being drawn in. It is desirable

shown, a similar symmetrical curve

in plotting this curve to draw, as
in the third quadrant joining the real curve in the first quadrant to
the imaginary curve in the third quadrant at the origin where there
isa point of inflection. This facilitates drawing a curve which has
the proper direction at the origin. Then drawing the tangent at
the origin we determine the line for 7/= CfR, and taking at any

point the simultaneous values of 7/ and R we have C/=

Another method is to plot a curve of 7 divided by R in the first

quadrant and a symmetrical curve in the second quadrant. Such

a curve will not pass through the origin but cut the axis of R = zero
at a point above the origin. Its ordinate here is evidently C/. The
ordinates of the curve of bear a constant ratio to the ordinates of
the curve of mean effective pressure.
It is
customary to reduce the initial friction or internal resistance
of an engine to equivalent mean effective pressure in the low pressure
cylinder or cylinders. This is the most convenient and probably
the most reliable way of comparing engines of different types and
sizes as regards internal resistance.

Let n denote the number of L.P. cylinders, d the diameter of

each in inches, 5 the stroke in inches, p m the mean effective pressure
in pounds per square inch reduced to the L.P. cylinder area and R
the revolutions per minute. Then
7T 2 S
_ 4 12


At the limit 7 = //
= CfR. If pf denote mean
effective pressure

equivalent to internal resistance reduced to L.P. area, at the limit

p m = pf or
Cf = -nd?spf or pf =

252100 nsd?

It is seen from the above that when we have once determined a

reliable value of C/ we can readily obtain the corresponding value
of the mean effective pressure in the low pressure cylinder from the
known data of the engine. If we could determine with accuracy
the curve of indicated horse-power for a given engine to a very low
number of revolutions the above method of determining internal
resistance would leave little to be desired. However, we meet here
with a number of practical difficulties. If we determine simulta-
neous values of speed, power and revolutions, which is the usual
practice in progressive trials, it is found that the low speed trials
over a measured mile are very tedious. If we avoid this trouble by
determining in free route at the lowest speeds the horse-power and
revolutions only, we still encounter difficulties. No reciprocating
engine will run at all below a certain speed, and as it approaches
the limiting speed at which it will stick, its action becomes some-
what erratic and uncertain. It is true that the less the friction the
lower the revolutions at which the engine will stick, and that this
is a rough measure of the initial friction; but even the smoothest

running engines will seldom run steadily down to a speed sufficiently

low to enable the internal resistance to be determined with accuracy
by a curve extended to the origin. For determining very low speed
powers of engines which use high pressure it is necessary to use
special weak
indicator springs, otherwise the indicator diagrams
have such a very small area that the determination of the power is
very uncertain. If, instead of determining a curve of power and
revolutions for the ship under way we determine the same thing for
the vessel tiedup at the dock, we will get larger indicator cards and
the engine will turn over at a slightly lower number of revolutions,
but even then the results generally leave something to be desired.
Torsion meter apparatus has been designed of late years to
measure the power being transmitted by a shaft by determining
the twist of the shaft. If we measure shaft horse-power by a tor-

sion meter and simultaneously indicate the engine, we can determine

the total frictional resistance of the engine, the power absorbed by
friction being of course the difference between the indicated horse-

power and the shaft horse-power. With accurate data this would
probably be the most nearly exact method of determining the initial
friction of the engine and would have the incidental advantage of

enabling the load friction to be determined as well, but the accuracy

of torsion meters at low speeds and powers is not sufficient to enable
this method to be made use of except perhaps in very exceptional
cases. It is evident that we need some method of obtaining the
desired result from an ordinary curve of power and revolutions which
does not go below a speed and power for which the data may be
readily obtained and regarded as fairly reliable. It is natural to
ask whether there any inherent feature or property of curves of

horse-power which would facilitate the determination of the internal

friction. Mr. William Froude worked on these lines. He plotted
a curve of indicated thrust upon the speed of the ship in knots,
carrying the curve down as low as possible. Indicated thrust is a
thrust which at the speed of the propeller will absorb the indicated

horse-power. At zero speed and zero revolutions the curve of indi-

cated thrust, whose ordinates are proportional to will cut the

R ,

axis of thrust at a distance above the origin proportional to the

initial friction. To pass from the lowest point of his curve of indi-
cated thrust, determined by observation, Mr. Froude made use of
an essential property of these curves. He assumed that at these
low speeds the resistance of the ship varied as the 1.87 power of the

speed, and that all other losses except the initial friction loss were
constant fractions of the power absorbed by resistance. It would
follow that the curve of indicated thrust in the vicinity of the origin
is a parabola of the 1.87 degreewhose ordinate at zero speed is

proportional to the initial friction.

Now, referring to Fig. 262, if the curve therein indicated is a
parabola of the 1.87 degree it follows that the tangent at the point
P will cut the horizontal tangent through the lowest point A at a.

point M, so that AM divided by A N is equal to- Mr. Froude,

then, having drawn his curve of indicated thrust to as low a speed

as he could from the data, next drew the tangent at its extremity
OTJ 87
as KB in Fig. 263, and dividing H so that 7
OL at equals
OL 1.87
he set up HB to intersect the tangent at K in the point B. A hori-
zontal line, then, through B
cuts the axis of thrust at the point 7

and Or is the indicated thrust corresponding to the initial friction.

This method makes use of a property of the curve, but as a matter
it is hardly so reliable in practice as the method of extending
of fact,
the curve of indicated horse-power to the origin and setting off the
tangent to it. While the low speed resistance of the ship would be
reasonably close to the 1.87 power of the speed this is still an approxi-
mation, but the principal objection to this method is that it requires
a tangent to be drawn at the low speed extremity of the curve of
indicated thrust. The difficulty of obtaining reliable values for this
curve at the lowest speed have been pointed out and it follows,
apart from the difficulty of drawing an accurate tangent at the
extremity of any curve, that an error in the low speed spot would
throw out the low speed tangent and introduce material errors.

3. Initial Friction Deduced from Low Speed Portion of Power

Curves. The question arises, then, whether we cannot make use of
some inherent property of curves of horse-power which will enable
us to determine the with reasonable accuracy without
initial friction

it being necessary to carry any curve to the origin. We know that

the frictional resistance of a ship varies about as the 1.83 power of
the speed, so that the horse-power absorbed by frictional resistance
varies as the 2.83 power of the speed. The power absorbed by
wave making varies as a higher power than the cube of the speed.
The practical result is that at low speeds, when there is almost
no wave resistance, the total effective horse-power will vary as a
somewhat lower power of the speed than the cube, whereas at high
speeds it will vary as a higher power of the speed than the cube.
There is then some point at moderate speed where the effective
horse-power is varying as the cube of the speed.
Consider now the propeller. For a given slip the power absorbed
by a propeller varies as the cube of the revolutions, or for constant
slip as the cube of the speed. It follows, then, that starting from a

very low speed, where the effective horse-power is varying at a lower

power than the cube, the slip of the propeller falls off until we reach
the speed at which the effective horse-power varies as the cube of
the speed. At this point the slip of the propeller reaches a minimum
beyond which it increases. The efficiency of the propeller at the
point where the slip reaches a minimum
will be constant, and the

power delivered will vary as the cube of the speed or as the cube of
the revolutions. Also all losses will vary as the power delivered to
the propeller or as the cube of the revolutions, except the initial

friction loss.Hence, at the point of minimum slip where the slip

remains constant for a minute interval the following formula will

express exactly the indicated horse-power:

For some little distance on either side of the point of minimum

above formula will give a reasonably close approximation
slip the
to the facts, especially for the speeds below the point of minimum

slip. Now Cy and c in the above equation are both unknown, but
from the curve of indicated horse-power plotted on revolutions we
can determine any number of simultaneous values of 7 and R, and
for each pair of such values we can draw a straight line on axes of c
and Cf, constituting a focal diagram. If the equation above applies

throughout to the curve of indicated horse-power and C/ and c were

constant, it would follow that this diagram would have a perfect
focus. Now we know that the above equation does not apply to the
upper part of the curve of horse-power at which the indicated horse-
power generally varies as a very much higher power than the cube.
It seems reasonable from the nature of the case, however, that this

equation should be fairly approximate over a tolerably wide range

of the lower speeds,and that if we draw for this range a series of
lines Cf and they should all pass reasonably close to a common

point; in other words, should form a reliable focal diagram. Inves-

tigation of practical cases shows that we do have such a focal
diagram. The methods of calculation are very simple. The
table below shows the calculations for the Yorktown, and Fig. 264
shows the diagram for the Yorktown.




The above vessels were all given carefultrials and the results are

as reliable as will usually be obtained. While the diagrams show

lines for successive speeds, successive values of revolutions could
have been used as well, and in fact the method can be readily applied
to a curve of power and revolutions where the speed is not known.
It is seen that in every case there is an excellent focus formed by

the lines for the lower speeds, except in the case of the Maine,
where the focus is not so well defined as would be desirable. The
generally satisfactory determination of the focus in accordance with
theoretical reasoning may be regarded as fairly strong evidence in
favor of the method outlined above. To produce direct evidence
for thismethod we can apply it to a case where the internal resis-
tance is accurately known by some other method. The Yorktown
was one such case. Fortunately, however, we can produce stronger
cases. In the transactions of the Society of Naval Architects and
Marine Engineers we find two cases of determinations of speed and
power of double-ended ferry boats with a propeller at each end.
Three curves are given for each case, one curve for both screws in
use,one for only the stern screw in use, the bow screw being removed,
and one for only the bow screw in use. One case was that of the
Cincinnati, the data for which can be found in a paper by F. L.
DuBosque, in the volume for 1896, and the other case was that of
the Edgewaier, the data for which can be found in a paper by
E. A. Stevens in the volume for 1902. Fig. 270 reproduces the
curves of power plotted on revolutions for the Cincinnati and
Fig. 271 the similar curves for the Edgewaier. It is seen that the
three curves for each boat differ radically from each other, owing
to differences of propeller efficiencies, etc., but it is evident that for
each vessel the internal friction of the engine should not vary much
for the three conditions, since the engines, shafting, etc., were the
same and the only factors affecting frictional resistance were the

presence or absence of one screw and the variations of initial friction

between trials. Figures 272 and 273 show the frictional focal dia-
grams for the Cincinnati and Edgewaier as deduced from Figs. 270
and 271 and the curves of speed and revolutions. The original obser-
vations for the Cincinnati do not extend to quite so low a speed as
desirable for the initial friction determination, but it is seen that the

several cases, in spite of the radical differences in the curves of power,

give fairly satisfactory fo